Package 'mlr3extralearners'

Title: Extra Learners For mlr3
Description: Extra learners for use in mlr3.
Authors: Raphael Sonabend [aut] (ORCID: <https://orcid.org/0000-0001-9225-4654>), Patrick Schratz [aut] (ORCID: <https://orcid.org/0000-0003-0748-6624>), Lorenz A. Kapsner [ctb] (ORCID: <https://orcid.org/0000-0003-1866-860X>), Lennart Schneider [ctb] (ORCID: <https://orcid.org/0000-0003-4152-5308>), Stephen A Lauer [ctb] (ORCID: <https://orcid.org/0000-0003-2948-630X>), Pierre Camilleri [ctb], Javier García [ctb], Keenan Ganz [ctb], Byron Jaeger [ctb], Sebastian Fischer [cre, aut], Zezhi Wang [ctb], John Zobolas [aut] (ORCID: <https://orcid.org/0000-0002-3609-8674>), Lukas Burk [ctb] (ORCID: <https://orcid.org/0000-0001-7528-3795>), Lona Koers [ctb], Anna Nazarova [ctb], Baisu Zhou [ctb], Marc Becker [ctb] (ORCID: <https://orcid.org/0000-0002-8115-0400>), Nikolai German [ctb], Alexander Winterstetter [ctb], Imad El Badisy [ctb]
Maintainer: Sebastian Fischer <[email protected]>
License: LGPL-3
Version: 1.5.2
Built: 2026-05-22 19:00:35 UTC
Source: https://github.com/mlr-org/mlr3extralearners

Help Index

mlr3extralearners: Extra Learners For mlr3

Description

Extra learners for use in mlr3.

Author(s)

Maintainer: Sebastian Fischer [email protected]

Authors:

Other contributors:

See Also

Useful links:

Assert Python Packages

Description

Assert that the given Python packages are available.

Usage

assert_python_packages(packages, python_version = NULL)

Arguments

packages

(character())
Python packages to check.
python_version

(character(1))
Python version to use. If NULL, the default Python version is used.

Value

(character())

Install Learner Dependencies

Description

Install required dependencies for specified learners. Works for packages on GitHub and CRAN, otherwise must be manually installed.

Usage

install_learners(.keys, repos = "https://cloud.r-project.org", ...)

Arguments

.keys

(character())
Keys passed to mlr_learners specifying learners to install.
repos

(character(1))
Passed to utils::install.packages.
...

(ANY)
Additional options to pass to utils::install.packages() or remotes::install_github().

List Learners in mlr3verse

Description

Lists all learners, properties, and associated packages in a table that can be filtered and queried.

Usage

list_mlr3learners(select = NULL, filter = NULL)

Arguments

select

character()
Passed to data.table::subset.
filter

list()
Named list of conditions to filter on, names correspond to column names in table.

Classification Abess Learner

Description

Adaptive best-subset selection for classification. Calls abess::abess() from abess.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.abess")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, abess

Parameters

Id Type Default Levels Range
family character - binomial, multinomial, ordinal -
tune.path character sequence sequence, gsection -
tune.type character gic gic, aic, bic, ebic, cv -
normalize integer NULL (−∞,∞)(-\infty, \infty)
support.size untyped NULL -
c.max integer 2 [1,∞)[1, \infty)
gs.range untyped NULL -
lambda numeric 0 [0,∞)[0, \infty)
always.include untyped NULL -
group.index untyped NULL -
init.active.set untyped NULL -
splicing.type integer 2 [1,2][1, 2]
max.splicing.iter integer 20 [1,∞)[1, \infty)
screening.num integer NULL [0,∞)[0, \infty)
important.search integer NULL [0,∞)[0, \infty)
warm.start logical TRUE TRUE, FALSE -
nfolds integer 5 (−∞,∞)(-\infty, \infty)
foldid untyped NULL -
cov.update logical FALSE TRUE, FALSE -
newton character exact exact, approx -
newton.thresh numeric 1e-06 [0,∞)[0, \infty)
max.newton.iter integer NULL [1,∞)[1, \infty)
early.stop logical FALSE TRUE, FALSE -
ic.scale numeric 1 [0,∞)[0, \infty)
num.threads integer 0 [0,∞)[0, \infty)
seed integer 0 (−∞,∞)(-\infty, \infty)

Initial parameter values

  • num.threads: This parameter is initialized to 1 (default is 0) to avoid conflicts with the mlr3 parallelization.

  • family: Depends on the task type, if the parameter family is NULL, it is set to "binomial" for binary classification tasks and to "multinomial" for multiclass classification problems.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifAbess

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifAbess$new()

Method selected_features()

Extract the name of selected features from the model by abess::extract().

Usage
LearnerClassifAbess$selected_features()
Returns

The names of selected features

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifAbess$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bbayukari

See Also

Examples

# Define the Learner
learner = lrn("classif.abess")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Boosting Learner

Description

Classification using ada. Calls ada::ada() from ada.

Initial parameter values

  • xval is set to 0 by default for speed.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.ada")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, ada, rpart

Parameters

Id Type Default Levels Range
bag.frac numeric 0.5 [0,1][0, 1]
bag.shift logical FALSE TRUE, FALSE -
cp numeric 0.01 [0,1][0, 1]
delta numeric 1e-10 [0,∞)[0, \infty)
iter integer 50 [1,∞)[1, \infty)
loss character exponential exponential, logistic -
max.iter integer 20 [1,∞)[1, \infty)
maxcompete integer 4 [0,∞)[0, \infty)
maxdepth integer 30 [1,30][1, 30]
maxsurrogate integer 5 [0,∞)[0, \infty)
minbucket integer - [1,∞)[1, \infty)
minsplit integer 20 [1,∞)[1, \infty)
model.coef logical TRUE TRUE, FALSE -
n.iter integer 50 [1,∞)[1, \infty)
nu numeric 0.1 [0,∞)[0, \infty)
surrogatestyle integer 0 [0,1][0, 1]
type character discrete discrete, real, gentle -
usesurrogate integer 2 [0,2][0, 2]
verbose logical FALSE TRUE, FALSE -
xval integer 0 [0,∞)[0, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifAdaBoosting

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifAdaBoosting$new()

Method oob_error()

The oob error is extracted extracted from the model slot ⁠$model$errs⁠.

Usage
LearnerClassifAdaBoosting$oob_error()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifAdaBoosting$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

See Also

Examples

# Define the Learner
learner = lrn("classif.ada")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Boosting Learner

Description

Classification boosting algorithm. Calls adabag::boosting() from adabag.

Initial parameter values

  • xval:

    • Actual default: 10L

    • Initial value: 0L

    • Reason for change: Set to 0 for speed.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.adabag")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, adabag, rpart

Parameters

Id Type Default Levels Range
boos logical TRUE TRUE, FALSE -
coeflearn character Breiman Breiman, Freund, Zhu -
cp numeric 0.01 [0,1][0, 1]
maxcompete integer 4 [0,∞)[0, \infty)
maxdepth integer 30 [1,30][1, 30]
maxsurrogate integer 5 [0,∞)[0, \infty)
mfinal integer 100 [1,∞)[1, \infty)
minbucket integer - [1,∞)[1, \infty)
minsplit integer 20 [1,∞)[1, \infty)
newmfinal integer - (−∞,∞)(-\infty, \infty)
surrogatestyle integer 0 [0,1][0, 1]
usesurrogate integer 2 [0,2][0, 2]
xval integer 0 [0,∞)[0, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifAdabag

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifAdabag$new()

Method importance()

The importance scores are extracted from the model.

Usage
LearnerClassifAdabag$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifAdabag$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Alfaro, Esteban, Gamez, Matias, García, Noelia (2013). “adabag: An R Package for Classification with Boosting and Bagging.” Journal of Statistical Software, 54(2), 1-35. doi:10.18637/jss.v054.i02, https://www.jstatsoft.org/index.php/jss/article/view/v054i02.

See Also

Examples

# Define the Learner
learner = lrn("classif.adabag", mfinal = 10L)
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification AdaBoostM1 Learner

Description

Adaptive boosting algorithm for classification. Calls RWeka::AdaBoostM1() from RWeka.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.AdaBoostM1")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
P integer 100 [90,100][90, 100]
Q logical FALSE TRUE, FALSE -
S integer 1 [1,∞)[1, \infty)
I integer 10 [1,∞)[1, \infty)
W untyped "DecisionStump" -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Parameter changes

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifAdaBoostM1

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifAdaBoostM1$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifAdaBoostM1$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifAdaBoostM1$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifAdaBoostM1$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

henrifnk

References

Freund, Yoav, Schapire, E R, others (1996). “Experiments with a new boosting algorithm.” In icml, volume 96, 148–156. Citeseer.

See Also

Examples

# Define the Learner
learner = lrn("classif.AdaBoostM1")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Oblique Random Forest Classifier

Description

Accelerated oblique random classification forest. Calls aorsf::orsf() from aorsf. Note that although the learner has the property "missing" and it can in principle deal with missing values, the behaviour has to be configured using the parameter na_action.

Initial parameter values

  • n_thread: This parameter is initialized to 1 (default is 0) to avoid conflicts with the mlr3 parallelization.

  • pred_simplify has to be TRUE, otherwise response is NA in prediction

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.aorsf")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, aorsf

Parameters

Id Type Default Levels Range
attach_data logical TRUE TRUE, FALSE -
epsilon numeric 1e-09 [0,∞)[0, \infty)
importance character anova none, anova, negate, permute -
importance_max_pvalue numeric 0.01 [1e−04,0.9999][1e-04, 0.9999]
leaf_min_events integer 1 [1,∞)[1, \infty)
leaf_min_obs integer 5 [1,∞)[1, \infty)
max_iter integer 20 [1,∞)[1, \infty)
method character glm glm, net, pca, random -
mtry integer NULL [1,∞)[1, \infty)
mtry_ratio numeric - [0,1][0, 1]
n_retry integer 3 [0,∞)[0, \infty)
n_split integer 5 [1,∞)[1, \infty)
n_thread integer - [0,∞)[0, \infty)
n_tree integer 500 [1,∞)[1, \infty)
na_action character fail fail, impute_meanmode -
net_mix numeric 0.5 (−∞,∞)(-\infty, \infty)
oobag logical FALSE TRUE, FALSE -
oobag_eval_every integer NULL [1,∞)[1, \infty)
oobag_fun untyped NULL -
oobag_pred_type character prob none, leaf, prob, class -
pred_aggregate logical TRUE TRUE, FALSE -
sample_fraction numeric 0.632 [0,1][0, 1]
sample_with_replacement logical TRUE TRUE, FALSE -
scale_x logical FALSE TRUE, FALSE -
split_min_events integer 5 [1,∞)[1, \infty)
split_min_obs integer 10 [1,∞)[1, \infty)
split_min_stat numeric NULL [0,∞)[0, \infty)
split_rule character gini gini, cstat -
target_df integer NULL [1,∞)[1, \infty)
tree_seeds integer NULL [1,∞)[1, \infty)
verbose_progress logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifObliqueRandomForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifObliqueRandomForest$new()

Method oob_error()

OOB concordance error extracted from the model slot eval_oobag$stat_values

Usage
LearnerClassifObliqueRandomForest$oob_error()
Returns

numeric().

Method importance()

The importance scores are extracted from the model.

Usage
LearnerClassifObliqueRandomForest$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifObliqueRandomForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

See Also

Examples

# Define the Learner
learner = lrn("classif.aorsf")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification BART (Bayesian Additive Regression Trees) Learner

Description

Bayesian Additive Regression Trees are similar to gradient boosting algorithms. The classification problem is solved by 0-1 encoding of the two-class targets and setting the decision threshold to p = 0.5 during the prediction phase. Calls dbarts::bart() from dbarts.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.bart")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, dbarts

Parameters

Id Type Default Levels Range
ntree integer 200 [1,∞)[1, \infty)
k numeric 2 [0,∞)[0, \infty)
power numeric 2 [0,∞)[0, \infty)
base numeric 0.95 [0,1][0, 1]
binaryOffset numeric 0 (−∞,∞)(-\infty, \infty)
ndpost integer 1000 [1,∞)[1, \infty)
nskip integer 100 [0,∞)[0, \infty)
printevery integer 100 [0,∞)[0, \infty)
keepevery integer 1 [1,∞)[1, \infty)
keeptrainfits logical TRUE TRUE, FALSE -
usequants logical FALSE TRUE, FALSE -
numcut integer 100 [1,∞)[1, \infty)
printcutoffs integer 0 (−∞,∞)(-\infty, \infty)
verbose logical FALSE TRUE, FALSE -
nthread integer 1 (−∞,∞)(-\infty, \infty)
keepcall logical TRUE TRUE, FALSE -
sampleronly logical FALSE TRUE, FALSE -
seed integer NA (−∞,∞)(-\infty, \infty)
proposalprobs untyped NULL -
splitprobs untyped NULL -
keepsampler logical - TRUE, FALSE -
n.threads integer - (−∞,∞)(-\infty, \infty)

Parameter Changes

  • Parameter: keeptrees

  • Original: FALSE

  • New: TRUE

  • Reason: Required for prediction

  • Parameter: offset

  • The parameter is removed, because only dbarts::bart2 allows an offset during training, and therefore the offset parameter in dbarts:::predict.bart is irrelevant for dbarts::dbart.

  • Parameter: nchain, combineChains, combinechains

  • The parameters are removed as parallelization of multiple models is handled by future.

  • Parameter: sigest, sigdf, sigquant, keeptres

  • Regression only.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifBart

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifBart$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifBart$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

ck37

References

Sparapani, Rodney, Spanbauer, Charles, McCulloch, Robert (2021). “Nonparametric machine learning and efficient computation with bayesian additive regression trees: the BART R package.” Journal of Statistical Software, 97, 1–66.

Chipman, A H, George, I E, McCulloch, E R (2010). “BART: Bayesian additive regression trees.” The Annals of Applied Statistics, 4(1), 266–298.

See Also

Examples

# Define the Learner
learner = lrn("classif.bart")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Bayes Network Learner

Description

Bayes Network learning using various search algorithms. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

  • N removed:

    • Initial structure is empty

  • P removed:

    • Maximum number of parents

  • R removed:

    • Random order

  • mbc removed:

    • Applies a Markov Blanket correction to the network structure, after a network structure is learned

  • S removed:

    • Score type

  • A removed:

    • Initial count (alpha)

  • Reason for change: The parameters are removed because they don't work out of the box and it's unclear how to use them.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.bayes_net")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
D logical - TRUE, FALSE -
B untyped - -
Q character - global.K2, global.HillClimber, global.SimulatedAnnealing, global.TabuSearch, global.TAN, local.K2, local.HillClimber, local.LAGDHillClimber, local.SimulatedAnnealing, local.TabuSearch, [...] -
E character - estimate.SimpleEstimator, estimate.BMAEstimator, estimate.MultiNomialBMAEstimator -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifBayesNet

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifBayesNet$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifBayesNet$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifBayesNet$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifBayesNet$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("classif.bayes_net")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

BlockForest Classification Learner

Description

Random forests for blocks of clinical and omics covariate data. Calls blockForest::blockfor() from package blockForest.

In this learner, only the trained forest object (⁠$forest⁠) is retained. The optimized block-specific tuning parameters (paramvalues) and the biased OOB error estimate (biased_oob_error_donotuse) are discarded, as they are either not needed for downstream use or not reliable for performance estimation.

Initial parameter values

  • num.threads is initialized to 1 to avoid conflicts with parallelization via future.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.blockforest")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, blockForest

Parameters

Id Type Default Levels Range
blocks untyped - -
block.method character BlockForest BlockForest, RandomBlock, BlockVarSel, VarProb, SplitWeights -
num.trees integer 2000 [1,∞)[1, \infty)
mtry untyped NULL -
nsets integer 300 [1,∞)[1, \infty)
num.trees.pre integer 1500 [1,∞)[1, \infty)
splitrule character extratrees extratrees, gini -
always.select.block integer 0 [0,1][0, 1]
importance character - none, impurity, impurity_corrected, permutation -
num.threads integer - [1,∞)[1, \infty)
seed integer NULL (−∞,∞)(-\infty, \infty)
verbose logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifBlockForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifBlockForest$new()

Method importance()

The importance scores are extracted from the model slot variable.importance.

Usage
LearnerClassifBlockForest$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifBlockForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Hornung, R., Wright, N. M (2019). “Block Forests: Random forests for blocks of clinical and omics covariate data.” BMC Bioinformatics, 20(1), 1–17. doi:10.1186/s12859-019-2942-y, https://doi.org/10.1186/s12859-019-2942-y.

See Also

Examples

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# check task's features
task$feature_names

# partition features to 2 blocks
blocks = list(bl1 = 1:42, bl2 = 43:60)

# define learner
learner = lrn("classif.blockforest", blocks = blocks,
              importance = "permutation", nsets = 10, predict_type = "prob",
              num.trees = 50, num.trees.pre = 10, splitrule = "gini")

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# feature importance
learner$importance()

# Make predictions for the test observations
pred = learner$predict(task, row_ids = ids$test)
pred

# Score the predictions
pred$score()

Classification Gradient Boosting Learner

Description

Classification gradient boosting learner. Calls bst::bst() from bst.

Initial parameter values

  • Learner = "ls": Default base learner type

  • xval = 0: No cross-validation

  • maxdepth = 1: Maximum tree depth

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.bst")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “numeric”

  • Required Packages: mlr3, mlr3extralearners, bst, rpart

Parameters

Id Type Default Levels Range
center logical FALSE TRUE, FALSE -
coefir untyped NULL -
cost numeric 0.5 [0,1][0, 1]
cp numeric 0.01 [0,1][0, 1]
df integer 4 [1,∞)[1, \infty)
family character hinge hinge, hinge2, binom, thingeDC, tbinomDC, binomdDC, loss, clossR, clossRMM, clossMM -
f.init untyped NULL -
fk untyped NULL -
intercept logical TRUE TRUE, FALSE -
iter integer 1 [1,∞)[1, \infty)
Learner character ls ls, sm, tree -
maxdepth integer 1 [1,30][1, 30]
maxsurrogate integer 5 [0,∞)[0, \infty)
minbucket integer - [1,∞)[1, \infty)
minsplit integer 20 [1,∞)[1, \infty)
mstop integer 50 [1,∞)[1, \infty)
numsample integer 50 [1,∞)[1, \infty)
nu numeric 0.1 [0,1][0, 1]
q numeric - [0,1][0, 1]
qh numeric - [0,1][0, 1]
s numeric - [0,∞)[0, \infty)
sh numeric - [0,∞)[0, \infty)
start logical FALSE TRUE, FALSE -
surrogatestyle integer 0 [0,1][0, 1]
threshold character adaptive adaptive, fixed -
trace logical FALSE TRUE, FALSE -
trun logical FALSE TRUE, FALSE -
twinboost logical FALSE TRUE, FALSE -
twintype integer 1 [1,2][1, 2]
xselect.init untyped NULL -
xval integer 10 [0,∞)[0, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifBst

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifBst$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifBst$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

Only classification-appropriate loss functions are available for the family parameter.

Author(s)

annanzrv

See Also

Examples

# Define the Learner
learner = lrn("classif.bst")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification C5.0 Learner

Description

Decision Tree Algorithm. Calls C50::C5.0.formula() from C50.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.C50")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, C50

Parameters

Id Type Default Levels Range
trials integer 1 [1,∞)[1, \infty)
rules logical FALSE TRUE, FALSE -
costs untyped NULL -
subset logical TRUE TRUE, FALSE -
bands integer - [0,1000][0, 1000]
winnow logical FALSE TRUE, FALSE -
noGlobalPruning logical FALSE TRUE, FALSE -
CF numeric 0.25 [0,1][0, 1]
minCases integer 2 [0,∞)[0, \infty)
fuzzyThreshold logical FALSE TRUE, FALSE -
sample numeric 0 [0,0.999][0, 0.999]
seed integer - (−∞,∞)(-\infty, \infty)
earlyStopping logical TRUE TRUE, FALSE -
label untyped "outcome" -
na.action untyped "stats::na.pass" -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifC50

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifC50$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifC50$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

henrifnk

References

Quinlan, Ross J (2014). C4. 5: programs for machine learning. Elsevier.

Gradient Boosted Decision Trees Classification Learner

Description

Gradient boosting algorithm that also supports categorical data. Calls catboost::catboost.train() from package 'catboost'.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.catboost")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, catboost

Parameters

Id Type Default Levels Range
loss_function_twoclass character Logloss Logloss, CrossEntropy -
loss_function_multiclass character MultiClass MultiClass, MultiClassOneVsAll -
learning_rate numeric 0.03 [0.001,1][0.001, 1]
random_seed integer 0 [0,∞)[0, \infty)
l2_leaf_reg numeric 3 [0,∞)[0, \infty)
bootstrap_type character - Bayesian, Bernoulli, MVS, Poisson, No -
bagging_temperature numeric 1 [0,∞)[0, \infty)
subsample numeric - [0,1][0, 1]
sampling_frequency character PerTreeLevel PerTree, PerTreeLevel -
sampling_unit character Object Object, Group -
mvs_reg numeric - [0,∞)[0, \infty)
random_strength numeric 1 [0,∞)[0, \infty)
depth integer 6 [1,16][1, 16]
grow_policy character SymmetricTree SymmetricTree, Depthwise, Lossguide -
min_data_in_leaf integer 1 [1,∞)[1, \infty)
max_leaves integer 31 [1,∞)[1, \infty)
ignored_features untyped NULL -
one_hot_max_size untyped FALSE -
has_time logical FALSE TRUE, FALSE -
rsm numeric 1 [0.001,1][0.001, 1]
nan_mode character Min Min, Max -
fold_permutation_block integer - [1,256][1, 256]
leaf_estimation_method character - Newton, Gradient, Exact -
leaf_estimation_iterations integer - [1,∞)[1, \infty)
leaf_estimation_backtracking character AnyImprovement No, AnyImprovement, Armijo -
fold_len_multiplier numeric 2 [1.001,∞)[1.001, \infty)
approx_on_full_history logical TRUE TRUE, FALSE -
class_weights untyped - -
auto_class_weights character None None, Balanced, SqrtBalanced -
boosting_type character - Ordered, Plain -
boost_from_average logical - TRUE, FALSE -
langevin logical FALSE TRUE, FALSE -
diffusion_temperature numeric 10000 [0,∞)[0, \infty)
score_function character Cosine Cosine, L2, NewtonCosine, NewtonL2 -
monotone_constraints untyped - -
feature_weights untyped - -
first_feature_use_penalties untyped - -
penalties_coefficient numeric 1 [0,∞)[0, \infty)
per_object_feature_penalties untyped - -
model_shrink_rate numeric - (−∞,∞)(-\infty, \infty)
model_shrink_mode character - Constant, Decreasing -
target_border numeric - (−∞,∞)(-\infty, \infty)
border_count integer - [1,65535][1, 65535]
feature_border_type character GreedyLogSum Median, Uniform, UniformAndQuantiles, MaxLogSum, MinEntropy, GreedyLogSum -
per_float_feature_quantization untyped - -
classes_count integer - [1,∞)[1, \infty)
thread_count integer 1 [−1,∞)[-1, \infty)
task_type character CPU CPU, GPU -
devices untyped - -
logging_level character Silent Silent, Verbose, Info, Debug -
metric_period integer 1 [1,∞)[1, \infty)
train_dir untyped "catboost_info" -
model_size_reg numeric 0.5 [0,1][0, 1]
allow_writing_files logical FALSE TRUE, FALSE -
save_snapshot logical FALSE TRUE, FALSE -
snapshot_file untyped - -
snapshot_interval integer 600 [1,∞)[1, \infty)
simple_ctr untyped - -
combinations_ctr untyped - -
ctr_target_border_count integer - [1,255][1, 255]
counter_calc_method character Full SkipTest, Full -
max_ctr_complexity integer - [1,∞)[1, \infty)
ctr_leaf_count_limit integer - [1,∞)[1, \infty)
store_all_simple_ctr logical FALSE TRUE, FALSE -
final_ctr_computation_mode character Default Default, Skip -
verbose logical FALSE TRUE, FALSE -
ntree_start integer 0 [0,∞)[0, \infty)
ntree_end integer 0 [0,∞)[0, \infty)
early_stopping_rounds integer - [1,∞)[1, \infty)
eval_metric untyped - -
use_best_model logical - TRUE, FALSE -
iterations integer 1000 [1,∞)[1, \infty)

Installation

See https://catboost.ai/en/docs/concepts/r-installation.

Initial parameter values

  • logging_level:

    • Actual default: "Verbose"

    • Adjusted default: "Silent"

    • Reason for change: consistent with other mlr3 learners

  • thread_count:

    • Actual default: -1

    • Adjusted default: 1

    • Reason for change: consistent with other mlr3 learners

  • allow_writing_files:

    • Actual default: TRUE

    • Adjusted default: FALSE

    • Reason for change: consistent with other mlr3 learners

  • save_snapshot:

    • Actual default: TRUE

    • Adjusted default: FALSE

    • Reason for change: consistent with other mlr3 learners

Early stopping

Early stopping can be used to find the optimal number of boosting rounds. Set early_stopping_rounds to an integer value to monitor the performance of the model on the validation set while training. For information on how to configure the validation set, see the Validation section of mlr3::Learner.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifCatboost

Active bindings

internal_valid_scores

The last observation of the validation scores for all metrics. Extracted from model$evaluation_log

internal_tuned_values

Returns the early stopped iterations if early_stopping_rounds was set during training.

validate

How to construct the internal validation data. This parameter can be either NULL, a ratio, "test", or "predefined".

Methods

Public methods

Inherited methods

Method new()

Create a LearnerClassifCatboost object.

Usage
LearnerClassifCatboost$new()

Method importance()

The importance scores are calculated using catboost::catboost.get_feature_importance(), setting type = "FeatureImportance", returned for 'all'.

Usage
LearnerClassifCatboost$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifCatboost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sumny

References

Dorogush, Veronika A, Ershov, Vasily, Gulin, Andrey (2018). “CatBoost: gradient boosting with categorical features support.” arXiv preprint arXiv:1810.11363.

See Also

Examples

# Define the Learner
learner = lrn("classif.catboost", iterations = 10)
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Conditional Random Forest Learner

Description

A random forest based on conditional inference trees (partykit::ctree()). Calls partykit::cforest() from partykit.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.cforest")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, partykit, sandwich, coin

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
replace logical FALSE TRUE, FALSE -
fraction numeric 0.632 [0,1][0, 1]
mtry integer - [0,∞)[0, \infty)
mtryratio numeric - [0,1][0, 1]
applyfun untyped - -
cores integer NULL (−∞,∞)(-\infty, \infty)
trace logical FALSE TRUE, FALSE -
cluster untyped - -
scores untyped - -
teststat character quadratic quadratic, maximum -
splitstat character quadratic quadratic, maximum -
splittest logical FALSE TRUE, FALSE -
testtype character Univariate Bonferroni, MonteCarlo, Univariate, Teststatistic -
nmax untyped - -
pargs untyped - -
alpha numeric 0.05 [0,1][0, 1]
mincriterion numeric 0 [0,1][0, 1]
logmincriterion numeric 0 (−∞,∞)(-\infty, \infty)
minsplit integer 20 [1,∞)[1, \infty)
minbucket integer 7 [1,∞)[1, \infty)
minprob numeric 0.01 [0,1][0, 1]
stump logical FALSE TRUE, FALSE -
lookahead logical FALSE TRUE, FALSE -
MIA logical FALSE TRUE, FALSE -
nresample integer 9999 [1,∞)[1, \infty)
tol numeric 1.490116e-08 [0,∞)[0, \infty)
maxsurrogate integer 0 [0,∞)[0, \infty)
numsurrogate logical FALSE TRUE, FALSE -
maxdepth integer Inf [0,∞)[0, \infty)
multiway logical FALSE TRUE, FALSE -
splittry integer 2 [0,∞)[0, \infty)
intersplit logical FALSE TRUE, FALSE -
majority logical FALSE TRUE, FALSE -
caseweights logical TRUE TRUE, FALSE -
saveinfo logical FALSE TRUE, FALSE -
update logical FALSE TRUE, FALSE -
splitflavour character ctree ctree, exhaustive -
maxvar integer - [1,∞)[1, \infty)
OOB logical FALSE TRUE, FALSE -
simplify logical TRUE TRUE, FALSE -
scale logical TRUE TRUE, FALSE -
nperm integer 1 [0,∞)[0, \infty)
risk character loglik loglik, misclassification -
conditional logical FALSE TRUE, FALSE -
threshold numeric 0.2 (−∞,∞)(-\infty, \infty)

Custom mlr3 parameters

  • mtry:

    • This hyperparameter can alternatively be set via the added hyperparameter mtryratio as mtry = max(ceiling(mtryratio * n_features), 1). Note that mtry and mtryratio are mutually exclusive.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifCForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifCForest$new()

Method oob_error()

The importance scores are calculated using partykit::varimp().

The out-of-bag error, calculated using the OOB predictions from partykit.

Usage
LearnerClassifCForest$oob_error()
Returns

numeric(1).

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifCForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sumny

References

Hothorn T, Zeileis A (2015). “partykit: A Modular Toolkit for Recursive Partytioning in R.” Journal of Machine Learning Research, 16(118), 3905-3909. http://jmlr.org/papers/v16/hothorn15a.html.

Hothorn T, Hornik K, Zeileis A (2006). “Unbiased Recursive Partitioning: A Conditional Inference Framework.” Journal of Computational and Graphical Statistics, 15(3), 651–674. doi:10.1198/106186006x133933, https://doi.org/10.1198/106186006x133933.

See Also

Examples

# Define the Learner
learner = lrn("classif.cforest", ntree = 50L)
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Conditional Inference Tree Learner

Description

Classification Partition Tree where a significance test is used to determine the univariate splits. Calls partykit::ctree() from partykit.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.ctree")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, partykit, sandwich, coin

Parameters

Id Type Default Levels Range
teststat character quadratic quadratic, maximum -
splitstat character quadratic quadratic, maximum -
splittest logical FALSE TRUE, FALSE -
testtype character Bonferroni Bonferroni, MonteCarlo, Univariate, Teststatistic -
nmax untyped - -
alpha numeric 0.05 [0,1][0, 1]
mincriterion numeric 0.95 [0,1][0, 1]
logmincriterion numeric - (−∞,∞)(-\infty, \infty)
minsplit integer 20 [1,∞)[1, \infty)
minbucket integer 7 [1,∞)[1, \infty)
minprob numeric 0.01 [0,1][0, 1]
stump logical FALSE TRUE, FALSE -
lookahead logical FALSE TRUE, FALSE -
MIA logical FALSE TRUE, FALSE -
nresample integer 9999 [1,∞)[1, \infty)
tol numeric - [0,∞)[0, \infty)
maxsurrogate integer 0 [0,∞)[0, \infty)
numsurrogate logical FALSE TRUE, FALSE -
mtry integer Inf [0,∞)[0, \infty)
maxdepth integer Inf [0,∞)[0, \infty)
multiway logical FALSE TRUE, FALSE -
splittry integer 2 [0,∞)[0, \infty)
intersplit logical FALSE TRUE, FALSE -
majority logical FALSE TRUE, FALSE -
caseweights logical FALSE TRUE, FALSE -
maxvar integer - [1,∞)[1, \infty)
applyfun untyped - -
cores integer NULL (−∞,∞)(-\infty, \infty)
saveinfo logical TRUE TRUE, FALSE -
update logical FALSE TRUE, FALSE -
splitflavour character ctree ctree, exhaustive -
cluster untyped - -
scores untyped - -
doFit logical TRUE TRUE, FALSE -
maxpts integer 25000 (−∞,∞)(-\infty, \infty)
abseps numeric 0.001 [0,∞)[0, \infty)
releps numeric 0 [0,∞)[0, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifCTree

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifCTree$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifCTree$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sumny

References

Hothorn T, Zeileis A (2015). “partykit: A Modular Toolkit for Recursive Partytioning in R.” Journal of Machine Learning Research, 16(118), 3905-3909. http://jmlr.org/papers/v16/hothorn15a.html.

Hothorn T, Hornik K, Zeileis A (2006). “Unbiased Recursive Partitioning: A Conditional Inference Framework.” Journal of Computational and Graphical Statistics, 15(3), 651–674. doi:10.1198/106186006x133933, https://doi.org/10.1198/106186006x133933.

See Also

Examples

# Define the Learner
learner = lrn("classif.ctree")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Cross-Validated MCP- or SCAD-penalized Classification Learner

Description

MCP- or SCAD-penalized binary classification model with k-fold cross validation for choosing the regularization parameter lambda. Calls ncvreg::cv.ncvreg() from package ncvreg with family set to "binomial". Covariates are internally standardized by ncvreg prior to model fitting.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.cv_ncvreg")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, ncvreg

Parameters

Id Type Default Levels Range
penalty character MCP MCP, SCAD, lasso -
gamma numeric - [0,∞)[0, \infty)
alpha numeric 1 [0,1][0, 1]
lambda.min numeric - [0,∞)[0, \infty)
nlambda integer 100 [1,∞)[1, \infty)
lambda untyped - -
eps numeric 1e-04 (−∞,∞)(-\infty, \infty)
max.iter integer 10000 [1,∞)[1, \infty)
convex logical TRUE TRUE, FALSE -
dfmax integer - [0,∞)[0, \infty)
penalty.factor untyped - -
warn logical TRUE TRUE, FALSE -
returnX logical TRUE TRUE, FALSE -
cluster untyped - -
nfolds integer 10 (−∞,∞)(-\infty, \infty)
fold untyped - -
returnY logical FALSE TRUE, FALSE -
trace logical FALSE TRUE, FALSE -
pred_lambda numeric - (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifNCVreg

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifNCVreg$new()

Method selected_features()

Returns the set of selected features which have non-zero coefficients. Calls the internal coef.cv.ncvreg() function.

Usage
LearnerClassifNCVreg$selected_features(lambda = NULL)
Arguments
lambda

(numeric(1))
Custom lambda, defaults to the lambda with the minimum CV error.

Returns

(character()) vector of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifNCVreg$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Breheny P, Huang J (2011). “Coordinate descent algorithms for nonconvex penalized regression, with applications to biological feature selection.” Annals of Applied Statistics, 5(1), 232–253. ISSN 1932-6157, doi:10.1214/10-AOAS388.

See Also

Examples

# Define the Learner
learner = lrn("classif.cv_ncvreg")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Deep Neural Network with DBN Pretraining Learner

Description

Deep neural network with weights initialized by a deep belief network. Calls deepnet::dbn.dnn.train() from deepnet.

This learner works with tasks that have at least two features.

Initial parameter values

  • output is set to "softmax" to enable class probabilities.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.dbnDNN")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, deepnet

Parameters

Id Type Default Levels Range
hidden untyped - -
activationfun character - sigm, linear, tanh -
learningrate numeric - [0,∞)[0, \infty)
momentum numeric - [0,∞)[0, \infty)
learningrate_scale numeric - [0,∞)[0, \infty)
numepochs integer - [1,∞)[1, \infty)
batchsize integer - [1,∞)[1, \infty)
output character - sigm, linear, softmax -
hidden_dropout numeric - [0,1][0, 1]
visible_dropout numeric - [0,1][0, 1]
cd integer - [1,∞)[1, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifDbnDNN

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifDbnDNN$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifDbnDNN$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Rong X (2014). deepnet: Deep Learning Toolkit in R. R package version 0.2.1, https://CRAN.R-project.org/package=deepnet.

See Also

Examples

# Define the Learner
learner = lrn("classif.dbnDNN")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Decision Stump Learner

Description

Decision Stump Learner. Calls RWeka::DecisionStump() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.decision_stump")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifDecisionStump

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifDecisionStump$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifDecisionStump$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifDecisionStump$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifDecisionStump$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("classif.decision_stump")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Decision Table Learner

Description

Simple Decision Table majority classifier. Calls RWeka::make_Weka_classifier() from RWeka.

Initial parameter values

  • E:

    • Has only 2 out of 4 original evaluation measures : acc and auc with acc being the default

    • Reason for change: this learner should only contain evaluation measures appropriate for classification tasks

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • P_best:

    • original id: P

  • D_best:

    • original id: D

  • N_best:

    • original id: N

  • S_best:

    • original id: S

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.decision_table")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
S character BestFirst BestFirst, GreedyStepwise -
X integer 1 (−∞,∞)(-\infty, \infty)
E character acc acc, auc -
I logical - TRUE, FALSE -
R logical - TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
P_best untyped - -
D_best character 1 0, 1, 2 -
N_best integer - (−∞,∞)(-\infty, \infty)
S_best integer 1 (−∞,∞)(-\infty, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifDecisionTable

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifDecisionTable$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifDecisionTable$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifDecisionTable$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifDecisionTable$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Kohavi R (1995). “The Power of Decision Tables.” In 8th European Conference on Machine Learning, 174–189.

See Also

Examples

# Define the Learner
learner = lrn("classif.decision_table")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Diagonal Linear Discriminant Analysis Learner

Description

Diagonal Linear Discriminant Analysis classifier. Belongs to the family of Naive Bayes classifiers, where the distributions of each class are assumed to be multivariate normal and to share a common covariance matrix. Off-diagonal elements of the pooled sample covariance matrix are set to zero Calls sparsediscrim::lda_diag() from sparsediscrim.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.dlda")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, sparsediscrim

Parameters

Id Type Default
prior untyped NULL

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifDiagLda

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifDiagLda$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifDiagLda$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Dudoit S, Fridlyand J, Speed TP (2002). “Comparison of Discrimination Methods for the Classification of Tumors Using Gene Expression Data.” Journal of the American Statistical Association, 97(457), 77–87. doi:10.1198/016214502753479248.

See Also

Examples

# Define the Learner
learner = lrn("classif.dlda")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification MARS (Multivariate Adaptive Regression Splines) Learner

Description

This is an alternative implementation of MARS (Multivariate Adaptive Regression Splines). The classification problem is solved by 0-1 encoding of the two-class targets and setting the decision threshold to p = 0.5 during the prediction phase. MARS is trademarked and thus not used as the name. The name "earth" stands for "Enhanced Adaptive Regression Through Hinges".

Details

Methods for variance estimations are not yet implemented.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.earth")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, earth

Parameters

Id Type Default Levels Range
wp untyped NULL -
offset untyped NULL -
keepxy logical FALSE TRUE, FALSE -
trace character 0 0, .3, .5, 1, 2, 3, 4, 5 -
degree integer 1 [1,∞)[1, \infty)
penalty numeric 2 [−1,∞)[-1, \infty)
nk untyped NULL -
thresh numeric 0.001 (−∞,∞)(-\infty, \infty)
minspan numeric 0 [0,∞)[0, \infty)
endspan numeric 0 [0,∞)[0, \infty)
newvar.penalty numeric 0 [0,∞)[0, \infty)
fast.k integer 20 [0,∞)[0, \infty)
fast.beta integer 1 [0,1][0, 1]
linpreds untyped FALSE -
allowed untyped - -
pmethod character backward backward, none, exhaustive, forward, seqrep, cv -
nprune integer - [0,∞)[0, \infty)
nfold integer 0 [0,∞)[0, \infty)
ncross integer 1 [0,∞)[0, \infty)
stratify logical TRUE TRUE, FALSE -
varmod.method character none none, const, lm, rlm, earth, gam, power, power0, x.lm, x.rlm, [...] -
varmod.exponent numeric 1 (−∞,∞)(-\infty, \infty)
varmod.conv numeric 1 [0,1][0, 1]
varmod.clamp numeric 0.1 (−∞,∞)(-\infty, \infty)
varmod.minspan numeric -3 (−∞,∞)(-\infty, \infty)
Scale.y logical FALSE TRUE, FALSE -
Adjust.endspan numeric 2 (−∞,∞)(-\infty, \infty)
Auto.linpreds logical TRUE TRUE, FALSE -
Force.weights logical FALSE TRUE, FALSE -
Use.beta.cache logical TRUE TRUE, FALSE -
Force.xtx.prune logical FALSE TRUE, FALSE -
Get.leverages logical TRUE TRUE, FALSE -
Exhaustive.tol numeric 1e-10 (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifEarth

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifEarth$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifEarth$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

pkopper

References

Milborrow, Stephen, Hastie, T, Tibshirani, R (2014). “Earth: multivariate adaptive regression spline models.” R package version, 3(7).

Friedman, H J (1991). “Multivariate adaptive regression splines.” The annals of statistics, 19(1), 1–67.

See Also

Examples

# Define the Learner
learner = lrn("classif.earth")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Decision Tree Learner

Description

Evolutionary learning of globally optimal classification trees. Calls evtree::evtree() fromevtree.

Initial parameter values

pmutatemajor, pmutateminor, pcrossover, psplit, and pprune, are scaled internally to sum to 100.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifEvtree

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifEvtree$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifEvtree$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

See Also

Examples

# Define the Learner
learner = lrn("classif.evtree")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Logistic Regression with Exhaustive Search

Description

Logistic Regression, where the features used in the model are selected by running Exhaustive Search.

Calls ExhaustiveSearch::ExhaustiveSearch() from ExhaustiveSearch.

Initial parameter values

  • family:

    • Actual default: NULL

    • Adjusted default: "binomial"

    • Reason for change: To comply with mlr3 architecture, we differentiate between classification and regression learners.

  • nThreads:

    • Actual default: NULL

    • Adjusted default: 1

    • Reason for change: Suppressing the automatic internal parallelization if cv.folds > 0.

  • quietly:

    • Actual default: FALSE

    • Adjusted default: TRUE

    • Reason for change: Suppression of constant printing to console

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.exhaustive_search")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, ExhaustiveSearch

Parameters

Id Type Default Levels Range
family character - gaussian, binomial -
performanceMeasure character - MSE, AIC -
combsUpTo integer - [1,∞)[1, \infty)
nResults integer 5000 [1,∞)[1, \infty)
nThreads integer - [1,∞)[1, \infty)
testSetIDs integer - [1,∞)[1, \infty)
errorVal untyped -1L -
quietly logical - TRUE, FALSE -
checkLarge logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifExhaustiveSearch

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifExhaustiveSearch$new()

Method selected_features()

Extracts selected features of this learner.

Usage
LearnerClassifExhaustiveSearch$selected_features()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifExhaustiveSearch$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

ngerman

See Also

Examples

# define the learner
learner = lrn("classif.exhaustive_search", predict_type = "prob", combsUpTo = 3)

# define the task
# and subset to 3 features to speed up the example
tsk_sonar = tsk("sonar")$select(c("V1", "V2", "V3"))

# train the learner
learner$train(tsk_sonar)

# extract selected features
learner$selected_features()

# predict on training task
learner$predict(tsk_sonar)

Classification Neural Network Learner

Description

Simple and fast neural nets for tabular data classification. Calls fastai::tabular_learner() from package fastai.

Installation

The Python dependencies are automatically installed via reticulate::py_require(). See Installing Python Dependencies for more details. You can manually specify a virtual environment by calling reticulate::use_virtualenv() prior to calling the ⁠$train()⁠ function. In this virtual environment, the fastai package and its dependencies must be installed.

Saving a Learner

In order to save a lrn("classif.fastai") for later usage, it is necessary to call the ⁠$marshal()⁠ method on the Learner before writing it to disk, as the object will otherwise not be saved correctly. After loading a marshaled lrn("classif.fastai") into R again, you then need to call ⁠$unmarshal()⁠ to transform it into a useable state.

Initial parameter values

  • n_epoch: Needs to be set for fastai::fit() to work. If no value is given, it is set to 5.

  • eval_metric: Needs to be set for fastai::predict() to work. If no value is given, it is set to fastai::accuracy().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.fastai")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, fastai, reticulate

Parameters

Id Type Default Levels Range
act_cls untyped - -
bn_cont logical TRUE TRUE, FALSE -
bn_final logical FALSE TRUE, FALSE -
drop_last logical FALSE TRUE, FALSE -
embed_p numeric 0 [0,1][0, 1]
emb_szs untyped NULL -
n_epoch integer 5 [1,∞)[1, \infty)
eval_metric untyped - -
layers untyped - -
loss_func untyped - -
lr numeric 0.001 [0,∞)[0, \infty)
metrics untyped - -
n_out integer - (−∞,∞)(-\infty, \infty)
num_workers integer - (−∞,∞)(-\infty, \infty)
opt_func untyped - -
patience integer 1 [1,∞)[1, \infty)
pin_memory logical TRUE TRUE, FALSE -
procs untyped NULL -
ps untyped NULL -
shuffle logical FALSE TRUE, FALSE -
train_bn logical TRUE TRUE, FALSE -
wd integer - [0,∞)[0, \infty)
wd_bn_bias logical FALSE TRUE, FALSE -
use_bn logical TRUE TRUE, FALSE -
y_range untyped NULL -
bs integer 50 (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifFastai

Active bindings

internal_valid_scores

(named list() or NULL) The validation scores extracted from eval_protocol which itself is set by fitting the fastai::tab_learner. If early stopping is activated, this contains the validation scores of the model for the optimal n_epoch, otherwise the n_epoch for the final model.

internal_tuned_values

(named list() or NULL) If early stopping is activated, this returns a list with n_epoch, which is the last epoch that yielded improvement w.r.t. the patience, extracted by max(eval_protocol$epoch)+1

validate

(numeric(1) or character(1) or NULL) How to construct the internal validation data. This parameter can be either NULL, a ratio, "test", or "predefined".

marshaled

(logical(1)) Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifFastai$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifFastai$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifFastai$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifFastai$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Howard, Jeremy, Gugger, Sylvain (2020). “Fastai: A Layered API for Deep Learning.” Information, 11(2), 108. ISSN 2078-2489, doi:10.3390/info11020108.

See Also

Fast Nearest Neighbour Classification

Description

Fast Nearest Neighbour Classification. Calls FNN::knn() from FNN.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.fnn")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, FNN

Parameters

Id Type Default Levels Range
k integer 1 [1,∞)[1, \infty)
algorithm character kd_tree kd_tree, cover_tree, brute -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifFNN

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifFNN$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifFNN$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Boltz, Sylvain, Debreuve, Eric, Barlaud, Michel (2007). “kNN-based high-dimensional Kullback-Leibler distance for tracking.” In Eighth International Workshop on Image Analysis for Multimedia Interactive Services (WIAMIS'07), 16–16. IEEE.

See Also

Examples

# Define the Learner
learner = lrn("classif.fnn")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Generalized Additive Model Learner

Description

Generalized additive models. Calls mgcv::gam() from package mgcv with family set to binomial.

Multilabel classification is not implemented yet.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.gam")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, mgcv

Parameters

Id Type Default Levels Range
formula untyped - -
method character GCV.Cp GCV.Cp, GACV.Cp, REML, P-REML, ML, P-ML -
optimizer untyped c("outer", "newton") -
scale numeric 0 (−∞,∞)(-\infty, \infty)
select logical FALSE TRUE, FALSE -
knots untyped NULL -
sp untyped NULL -
min.sp untyped NULL -
H untyped NULL -
gamma numeric 1 [1,∞)[1, \infty)
paraPen untyped NULL -
G untyped NULL -
in.out untyped NULL -
drop.unused.levels logical TRUE TRUE, FALSE -
drop.intercept logical FALSE TRUE, FALSE -
nei untyped - -
nthreads integer 1 [1,∞)[1, \infty)
irls.reg numeric 0 [0,∞)[0, \infty)
epsilon numeric 1e-07 [0,∞)[0, \infty)
maxit integer 200 (−∞,∞)(-\infty, \infty)
trace logical FALSE TRUE, FALSE -
mgcv.tol numeric 1e-07 [0,∞)[0, \infty)
mgcv.half integer 15 [0,∞)[0, \infty)
rank.tol numeric 1.490116e-08 [0,∞)[0, \infty)
nlm untyped list() -
optim untyped list() -
newton untyped list() -
idLinksBases logical TRUE TRUE, FALSE -
scalePenalty logical TRUE TRUE, FALSE -
efs.lspmax integer 15 [0,∞)[0, \infty)
efs.tol numeric 0.1 [0,∞)[0, \infty)
scale.est character fletcher fletcher, pearson, deviance -
edge.correct logical FALSE TRUE, FALSE -
ncv.threads integer 1 [1,∞)[1, \infty)
block.size integer 1000 (−∞,∞)(-\infty, \infty)
unconditional logical FALSE TRUE, FALSE -

Formula

A gam formula specific to the task at hand is required for the formula parameter (see example and ?mgcv::formula.gam). Beware, if no formula is provided, a fallback formula is used that will make the gam behave like a glm (this behavior is required for the unit tests). Only features specified in the formula will be used, superseding columns with col_roles "feature" in the task.

Offset

If a Task contains a column with the offset role, it is automatically incorporated during training via the offset argument in mgcv::gam(). No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifGam

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifGam$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifGam$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

pierrecamilleri

References

Hastie, J T, Tibshirani, J R (2017). Generalized additive models. Routledge.

Wood, Simon (2012). “mgcv: Mixed GAM Computation Vehicle with GCV/AIC/REML smoothness estimation.”

Examples

# simple example
t = tsk("spam")$filter(1:1000)
l = lrn("classif.gam")
l$param_set$set_values(
  formula = type ~ s(george, k = 3) + s(charDollar, k = 3) + s(edu)
)
l$train(t)
l$model

Boosted Generalized Additive Classification Learner

Description

Fit a generalized additive classification model using a boosting algorithm. Calls mboost::gamboost() from mboost.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.gamboost")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, mboost

Parameters

Id Type Default Levels Range
baselearner character bbs bbs, bols, btree -
dfbase integer 4 (−∞,∞)(-\infty, \infty)
family character Binomial Binomial, AdaExp, AUC, custom -
custom.family untyped - -
link character logit logit, probit -
type character adaboost glm, adaboost -
mstop integer 100 (−∞,∞)(-\infty, \infty)
nu numeric 0.1 (−∞,∞)(-\infty, \infty)
risk character inbag inbag, oobag, none -
oobweights untyped NULL -
trace logical FALSE TRUE, FALSE -
stopintern untyped FALSE -
na.action untyped stats::na.omit -

Offset

If a Task contains a column with the offset role, it is automatically incorporated via the offset argument in mboost's training function. No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifGAMBoost

Methods

Public methods

Inherited methods

Method new()

Create a LearnerClassifGAMBoost object.

Usage
LearnerClassifGAMBoost$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifGAMBoost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Bühlmann, Peter, Yu, Bin (2003). “Boosting with the L 2 loss: regression and classification.” Journal of the American Statistical Association, 98(462), 324–339.

See Also

Examples

# Define the Learner
learner = lrn("classif.gamboost")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Gaussian Process Learner

Description

Gaussian process for classification. Calls kernlab::gausspr() from kernlab. Parameters sigma, degree, scale, offset and order are added to make tuning kpar easier. If kpar is provided then these new parameters are ignored. If none are provided then the default "automatic" is used for kpar.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.gausspr")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, kernlab

Parameters

Id Type Default Levels Range
scaled untyped TRUE -
kernel character rbfdot rbfdot, polydot, vanilladot, tanhdot, laplacedot, besseldot, anovadot, splinedot -
sigma numeric - (−∞,∞)(-\infty, \infty)
degree numeric - (−∞,∞)(-\infty, \infty)
scale numeric - (−∞,∞)(-\infty, \infty)
offset numeric - (−∞,∞)(-\infty, \infty)
order numeric - (−∞,∞)(-\infty, \infty)
kpar untyped "automatic" -
tol numeric 0.001 [0,∞)[0, \infty)
fit logical TRUE TRUE, FALSE -
na.action untyped na.omit -
coupler character minpair minpair, pkpd -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifGausspr

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifGausspr$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifGausspr$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Karatzoglou, Alexandros, Smola, Alex, Hornik, Kurt, Zeileis, Achim (2004). “kernlab-an S4 package for kernel methods in R.” Journal of statistical software, 11(9), 1–20.

See Also

Examples

# Define the Learner
learner = lrn("classif.gausspr")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Gradient Boosting Classification Learner

Description

Gradient Boosting Classification Algorithm. Calls gbm::gbm() from gbm.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.gbm")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, gbm

Parameters

Id Type Default Levels Range
distribution character bernoulli bernoulli, adaboost, huberized, multinomial -
n.trees integer 100 [1,∞)[1, \infty)
interaction.depth integer 1 [1,∞)[1, \infty)
n.minobsinnode integer 10 [1,∞)[1, \infty)
shrinkage numeric 0.001 [0,∞)[0, \infty)
bag.fraction numeric 0.5 [0,1][0, 1]
train.fraction numeric 1 [0,1][0, 1]
cv.folds integer 0 (−∞,∞)(-\infty, \infty)
keep.data logical FALSE TRUE, FALSE -
verbose logical FALSE TRUE, FALSE -
n.cores integer 1 (−∞,∞)(-\infty, \infty)
var.monotone untyped - -

Initial parameter values

  • keep.data is initialized to FALSE to save memory.

  • n.cores is initialized to 1 to avoid conflicts with parallelization through future.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifGBM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifGBM$new()

Method importance()

The importance scores are extracted by gbm::relative.influence() from the model.

Usage
LearnerClassifGBM$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifGBM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Friedman, H J (2002). “Stochastic gradient boosting.” Computational statistics & data analysis, 38(4), 367–378.

See Also

Examples

# Define the Learner
learner = lrn("classif.gbm")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Boosted Generalized Linear Classification Learner

Description

Fit a generalized linear classification model using a boosting algorithm. Calls mboost::glmboost() from mboost.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.glmboost")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, mboost

Parameters

Id Type Default Levels Range
family character Binomial Binomial, AdaExp, AUC, custom -
custom.family untyped - -
link character logit logit, probit -
type character adaboost glm, adaboost -
center logical TRUE TRUE, FALSE -
mstop integer 100 (−∞,∞)(-\infty, \infty)
nu numeric 0.1 (−∞,∞)(-\infty, \infty)
risk character inbag inbag, oobag, none -
oobweights untyped NULL -
trace logical FALSE TRUE, FALSE -
stopintern untyped FALSE -
na.action untyped stats::na.omit -
contrasts.arg untyped - -

Offset

If a Task contains a column with the offset role, it is automatically incorporated via the offset argument in mboost's training function. No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifGLMBoost

Methods

Public methods

Inherited methods

Method new()

Create a LearnerClassifGLMBoost object.

Usage
LearnerClassifGLMBoost$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifGLMBoost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Bühlmann, Peter, Yu, Bin (2003). “Boosting with the L 2 loss: regression and classification.” Journal of the American Statistical Association, 98(462), 324–339.

See Also

Examples

# Define the Learner
learner = lrn("classif.glmboost")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Linear Mixed Effect Learner

Description

Generalized linear model with random effects. Calls lme4::glmer() from lme4.

Initial parameter values

  • family - Is set to stats::binomial(link = "logit").

Formula

Although most mlr3 learners don't allow to specify the formula manually, and automatically set it by valling task$formula(), this learner allows to set the formula because it's core functionality depends it. This means that it might not always use all features that are available in the task. Be aware, that this can sometimes lead to unexpected error messages, because mlr3 checks the compatibility between the learner and the task on all available features.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.glmer")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, lme4

Parameters

Id Type Default Levels Range
formula untyped - -
start untyped NULL -
verbose integer 0 [0,∞)[0, \infty)
contrasts untyped NULL -
optimizer character - Nelder_Mead, bobyqa, nlminbwrap, nloptwrap -
restart_edge logical FALSE TRUE, FALSE -
boundary.tol numeric 1e-05 [0,∞)[0, \infty)
calc.derivs logical TRUE TRUE, FALSE -
autoscale untyped NULL -
check.nobs.vs.rankZ character ignore ignore, warning, message, stop -
check.nobs.vs.nlev character stop ignore, warning, message, stop -
check.nlev.gtreq.5 character ignore ignore, warning, message, stop -
check.nlev.gtr.1 character stop ignore, warning, message, stop -
check.nobs.vs.nRE character stop ignore, warning, message, stop -
check.rankX character message+drop.cols message+drop.cols, silent.drop.cols, warn+drop.cols, stop.deficient, ignore -
check.scaleX character warning warning, stop, silent.rescale, message+rescale, warn+rescale, ignore -
check.formula.LHS character stop ignore, warning, message, stop -
family untyped "stats::binomial(link = \"logit\")" -
nAGQ integer 1 [0,∞)[0, \infty)
mustart untyped - -
etastart untyped - -
check.conv.nobsmax numeric 10000 [1,∞)[1, \infty)
check.conv.nparmax numeric 20 [1,∞)[1, \infty)
check.conv.grad untyped "lme4::.makeCC(\"warning\", tol = 2e-3, relTol = NULL)" -
check.conv.singular untyped "lme4::.makeCC( action = \"message\", tol = formals(lme4::isSingular)$tol)" -
check.conv.hess untyped "lme4::.makeCC(action = \"warning\", tol = 1e-6)" -
optCtrl untyped list() -
tolPwrss untyped - -
compDev logical TRUE TRUE, FALSE -
nAGQ0initStep logical TRUE TRUE, FALSE -
check.response.not.const untyped "stop" -
newparams untyped NULL -
re.form untyped NULL -
random.only logical FALSE TRUE, FALSE -
allow.new.levels logical FALSE TRUE, FALSE -
na.action untyped "stats::na.pass" -

Offset

If a Task contains a column with the offset role, it is automatically incorporated during training via the offset argument in lme4::lmer(). No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifGlmer

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifGlmer$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifGlmer$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sebffischer

References

Bates, M D (2010). “lme4: Mixed-effects modeling with R.”

See Also

Examples

# Define the Learner and set parameter values
learner = lrn("classif.glmer",
  formula = credit_risk ~ (1 | credit_history) + job + property + telephone + savings)

# Define a Task
task = tsk("german_credit")
task$select(c("credit_history", "job", "property", "telephone", "savings"))

# Train the learner
learner$train(task)

print(learner$model)

Classification H2O Deep Learning Learner

Description

Classification feed-forward multilayer artificial neural network learner. Class h2o::h2o.deeplearning() from package h2o.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.h2o.deeplearning")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, h2o

Parameters

Id Type Default Levels Range
activation character Rectifier Rectifier, Tanh, TanhWithDropout, RectifierWithDropout, Maxout, MaxoutWithDropout -
adaptive_rate logical TRUE TRUE, FALSE -
auc_type character AUTO AUTO, NONE, MACRO_OVR, WEIGHTED_OVR, MACRO_OVO, WEIGHTED_OVO -
autoencoder logical FALSE TRUE, FALSE -
average_activation numeric 0 (−∞,∞)(-\infty, \infty)
balance_classes logical FALSE TRUE, FALSE -
categorical_encoding character AUTO AUTO, Enum, OneHotInternal, OneHotExplicit, Binary, Eigen, LabelEncoder, SortByResponse, EnumLimited -
checkpoint untyped NULL -
class_sampling_factors untyped NULL -
classification_stop numeric 0 [−1,∞)[-1, \infty)
diagnostics logical TRUE TRUE, FALSE -
distribution character AUTO AUTO, bernoulli, multinomial -
elastic_averaging logical FALSE TRUE, FALSE -
elastic_averaging_moving_rate numeric 0.9 (−∞,∞)(-\infty, \infty)
elastic_averaging_regularization numeric 0.001 (−∞,∞)(-\infty, \infty)
epochs numeric 10 [1,∞)[1, \infty)
epsilon numeric 1e-08 [1e−10,1e−04][1e-10, 1e-04]
export_checkpoints_dir untyped NULL -
export_weights_and_biases logical FALSE TRUE, FALSE -
fast_mode logical TRUE TRUE, FALSE -
force_load_balance logical TRUE TRUE, FALSE -
gainslift_bins integer -1 [−1,∞)[-1, \infty)
hidden untyped c(200L, 200L) -
hidden_dropout_ratios numeric 0.5 (−∞,∞)(-\infty, \infty)
ignore_const_cols logical TRUE TRUE, FALSE -
initial_weight_distribution character UniformAdaptive UniformAdaptive, Uniform, Normal -
initial_weight_scale numeric 1 (−∞,∞)(-\infty, \infty)
input_dropout_ratio numeric 0 (−∞,∞)(-\infty, \infty)
l1 numeric 0 (−∞,∞)(-\infty, \infty)
l2 numeric 0 (−∞,∞)(-\infty, \infty)
loss character Automatic Automatic, CrossEntropy, Quadratic -
max_after_balance_size numeric 5 (−∞,∞)(-\infty, \infty)
max_categorical_features integer NULL (−∞,∞)(-\infty, \infty)
max_runtime_secs numeric 0 [0,∞)[0, \infty)
max_w2 numeric 3.402823e+38 (−∞,∞)(-\infty, \infty)
mini_batch_size integer 1 (−∞,∞)(-\infty, \infty)
missing_values_handling character MeanImputation MeanImputation, Skip -
momentum_ramp numeric 1e+06 (−∞,∞)(-\infty, \infty)
momentum_stable numeric 0 (−∞,∞)(-\infty, \infty)
momentum_start numeric 0 (−∞,∞)(-\infty, \infty)
nesterov_accelerated_gradient logical TRUE TRUE, FALSE -
overwrite_with_best_model logical TRUE TRUE, FALSE -
pretrained_autoencoder untyped NULL -
quiet_mode logical TRUE TRUE, FALSE -
rate numeric 0.005 [0,1][0, 1]
rate_annealing numeric 1e-06 [0,∞)[0, \infty)
rate_decay numeric 1 [0,∞)[0, \infty)
replicate_training_data logical TRUE TRUE, FALSE -
reproducible logical FALSE TRUE, FALSE -
rho numeric 0.99 [0,∞)[0, \infty)
score_duty_cycle numeric 0.1 (−∞,∞)(-\infty, \infty)
score_each_iteration logical FALSE TRUE, FALSE -
score_interval numeric 5 (−∞,∞)(-\infty, \infty)
score_training_samples integer 10000 (−∞,∞)(-\infty, \infty)
score_validation_samples integer 0 (−∞,∞)(-\infty, \infty)
score_validation_sampling character Uniform Uniform, Stratified -
seed integer -1 (−∞,∞)(-\infty, \infty)
shuffle_training_data logical FALSE TRUE, FALSE -
single_node_mode logical FALSE TRUE, FALSE -
sparse logical FALSE TRUE, FALSE -
sparsity_beta numeric 0 (−∞,∞)(-\infty, \infty)
standardize logical TRUE TRUE, FALSE -
stopping_metric character AUTO AUTO, logloss, AUC, lift_top_group, misclassification, AUCPR, mean_per_class_error -
stopping_rounds integer 5 [0,∞)[0, \infty)
stopping_tolerance numeric 0 [0,∞)[0, \infty)
target_ratio_comm_to_comp numeric 0.05 (−∞,∞)(-\infty, \infty)
train_samples_per_iteration integer -2 [−2,∞)[-2, \infty)
verbose logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifH2ODeeplearning

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifH2ODeeplearning$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifH2ODeeplearning$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Fryda T, LeDell E, Gill N, Aiello S, Fu A, Candel A, Click C, Kraljevic T, Nykodym T, Aboyoun P, Kurka M, Malohlava M, Poirier S, Wong W (2025). h2o: R Interface for the 'H2O' Scalable Machine Learning Platform. R package version 3.46.0.9, https://github.com/h2oai/h2o-3.

See Also

Examples

# Define the Learner
learner = lrn("classif.h2o.deeplearning", epochs = 1)
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification H2O GBM Learner

Description

Gradient boosting machine learner. Class h2o::h2o.gbm() from package h2o.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.h2o.gbm")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, h2o

Parameters

Id Type Default Levels Range
auc_type character AUTO AUTO, NONE, MACRO_OVR, WEIGHTED_OVR, MACRO_OVO, WEIGHTED_OVO -
auto_rebalance logical TRUE TRUE, FALSE -
balance_classes logical FALSE TRUE, FALSE -
build_tree_one_node logical FALSE TRUE, FALSE -
categorical_encoding character AUTO AUTO, Enum, OneHotInternal, OneHotExplicit, Binary, Eigen, LabelEncoder, SortByResponse, EnumLimited -
check_constant_response logical TRUE TRUE, FALSE -
checkpoint untyped NULL -
class_sampling_factors untyped NULL -
col_sample_rate numeric 1 [0,1][0, 1]
col_sample_rate_change_per_level numeric 1 [0,2][0, 2]
col_sample_rate_per_tree numeric 1 [0,1][0, 1]
export_checkpoints_dir untyped NULL -
gainslift_bins integer -1 [−1,∞)[-1, \infty)
histogram_type character AUTO AUTO, UniformAdaptive, Random, QuantilesGlobal, RoundRobin, UniformRobust -
ignore_const_cols logical TRUE TRUE, FALSE -
in_training_checkpoints_dir untyped NULL -
in_training_checkpoints_tree_interval integer 1 [1,∞)[1, \infty)
interaction_constraints untyped NULL -
learn_rate numeric 0.1 [0,1][0, 1]
learn_rate_annealing numeric 1 [0,1][0, 1]
max_abs_leafnode_pred numeric Inf [0,∞)[0, \infty)
max_after_balance_size numeric 5 (−∞,∞)(-\infty, \infty)
max_depth integer 5 [1,∞)[1, \infty)
max_runtime_secs numeric 0 [0,∞)[0, \infty)
min_rows integer 10 [1,∞)[1, \infty)
min_split_improvement numeric 1e-05 [0,∞)[0, \infty)
monotone_constraints untyped NULL -
nbins integer 20 [1,∞)[1, \infty)
nbins_cats integer 1024 [1,∞)[1, \infty)
nbins_top_level integer 1024 [1,∞)[1, \infty)
ntrees integer 50 [1,∞)[1, \infty)
pred_noise_bandwidth numeric 0 [0,∞)[0, \infty)
sample_rate numeric 1 [0,1][0, 1]
sample_rate_per_class untyped NULL -
score_each_iteration logical FALSE TRUE, FALSE -
score_tree_interval integer 0 [0,∞)[0, \infty)
seed integer -1 (−∞,∞)(-\infty, \infty)
stopping_metric character AUTO AUTO, logloss, AUC, AUCPR, lift_top_group, misclassification, mean_per_class_error -
stopping_rounds integer 0 [0,∞)[0, \infty)
stopping_tolerance numeric 0.001 [0,∞)[0, \infty)
verbose logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifH2OGBM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifH2OGBM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifH2OGBM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Fryda T, LeDell E, Gill N, Aiello S, Fu A, Candel A, Click C, Kraljevic T, Nykodym T, Aboyoun P, Kurka M, Malohlava M, Poirier S, Wong W (2025). h2o: R Interface for the 'H2O' Scalable Machine Learning Platform. R package version 3.46.0.9, https://github.com/h2oai/h2o-3.

See Also

Examples

# Define the Learner
learner = lrn("classif.h2o.gbm")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification H2O GLM Learner

Description

Generalized linear classification model. Calls h2o::h2o.glm() from package h2o with family always set to "binomial".

H2O Connection

If no running H2O connection is found, the learner will automatically start a local H2O server on ⁠127.0.0.1⁠ via h2o::h2o.init(). If you want to connect to a remote H2O cluster, call h2o::h2o.init() with the appropriate arguments before training or predicting.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.h2o.glm")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, h2o

Parameters

Id Type Default Levels Range
alpha numeric 0.5 [0,1][0, 1]
balance_classes logical FALSE TRUE, FALSE -
beta_constraints untyped NULL -
beta_epsilon numeric 1e-04 [0,∞)[0, \infty)
build_null_model logical FALSE TRUE, FALSE -
calc_like logical FALSE TRUE, FALSE -
checkpoint untyped NULL -
class_sampling_factors untyped NULL -
cold_start logical FALSE TRUE, FALSE -
compute_p_values logical FALSE TRUE, FALSE -
early_stopping logical TRUE TRUE, FALSE -
export_checkpoints_dir untyped NULL -
gainslift_bins integer -1 [−1,∞)[-1, \infty)
generate_scoring_history logical FALSE TRUE, FALSE -
generate_variable_inflation_factors logical FALSE TRUE, FALSE -
gradient_epsilon numeric -1 [0,∞)[0, \infty)
HGLM logical FALSE TRUE, FALSE -
ignore_const_cols logical TRUE TRUE, FALSE -
interactions untyped NULL -
interaction_pairs untyped NULL -
intercept logical TRUE TRUE, FALSE -
lambda numeric 1e-05 [0,∞)[0, \infty)
lambda_min_ratio numeric -1 [0,1][0, 1]
lambda_search logical FALSE TRUE, FALSE -
link character logit family_default, logit -
max_active_predictors integer -1 [1,∞)[1, \infty)
max_after_balance_size numeric 5 [0,∞)[0, \infty)
max_iterations integer -1 [0,∞)[0, \infty)
max_runtime_secs numeric 0 [0,∞)[0, \infty)
missing_values_handling character MeanImputation MeanImputation, Skip, PlugValues -
nlambdas integer -1 [1,∞)[1, \infty)
non_negative logical FALSE TRUE, FALSE -
objective_epsilon numeric -1 [0,∞)[0, \infty)
obj_reg numeric -1 [0,∞)[0, \infty)
plug_values untyped NULL -
prior numeric -1 [0,∞)[0, \infty)
random_columns untyped NULL -
remove_collinear_columns logical FALSE TRUE, FALSE -
score_each_iteration logical FALSE TRUE, FALSE -
score_iteration_interval integer -1 (−∞,∞)(-\infty, \infty)
seed integer -1 (−∞,∞)(-\infty, \infty)
solver character AUTO AUTO, IRLSM, L_BFGS, COORDINATE_DESCENT, COORDINATE_DESCENT_NAIVE -
standardize logical TRUE TRUE, FALSE -
startval untyped NULL -
stopping_metric character AUTO AUTO, logloss, AUC, AUCPR, lift_top_group, misclassification, mean_per_class_error -
stopping_rounds integer 0 [0,∞)[0, \infty)
stopping_tolerance numeric 0.001 [0,∞)[0, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifH2OGLM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifH2OGLM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifH2OGLM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Fryda T, LeDell E, Gill N, Aiello S, Fu A, Candel A, Click C, Kraljevic T, Nykodym T, Aboyoun P, Kurka M, Malohlava M, Poirier S, Wong W (2025). h2o: R Interface for the 'H2O' Scalable Machine Learning Platform. R package version 3.46.0.9, https://github.com/h2oai/h2o-3.

See Also

Examples

# Define the Learner
learner = lrn("classif.h2o.glm")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification H2O Random Forest Learner

Description

Classification random forest learner. Calls h2o::h2o.randomForest() from package h2o.

H2O Connection

If no running H2O connection is found, the learner will automatically start a local H2O server on ⁠127.0.0.1⁠ via h2o::h2o.init(). If you want to connect to a remote H2O cluster, call h2o::h2o.init() with the appropriate arguments before training or predicting.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.h2o.randomForest")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, h2o

Parameters

Id Type Default Levels Range
auc_type character AUTO AUTO, NONE, MACRO_OVR, WEIGHTED_OVR, MACRO_OVO, WEIGHTED_OVO -
balance_classes logical FALSE TRUE, FALSE -
binomial_double_trees logical FALSE TRUE, FALSE -
build_tree_one_node logical FALSE TRUE, FALSE -
categorical_encoding character AUTO AUTO, Enum, OneHotInternal, OneHotExplicit, Binary, Eigen, LabelEncoder, SortByResponse, EnumLimited -
check_constant_response logical TRUE TRUE, FALSE -
checkpoint untyped NULL -
class_sampling_factors untyped NULL -
col_sample_rate_change_per_level numeric 1 [0,2][0, 2]
col_sample_rate_per_tree numeric 1 [0,1][0, 1]
export_checkpoints_dir untyped NULL -
gainslift_bins integer -1 [−1,∞)[-1, \infty)
histogram_type character AUTO AUTO, UniformAdaptive, Random, QuantilesGlobal, RoundRobin, UniformRobust -
ignore_const_cols logical TRUE TRUE, FALSE -
max_after_balance_size numeric 5 [0,∞)[0, \infty)
max_depth integer 20 [0,∞)[0, \infty)
max_runtime_secs numeric 0 [0,∞)[0, \infty)
min_rows numeric 1 [1,∞)[1, \infty)
min_split_improvement numeric 1e-05 [0,∞)[0, \infty)
mtries integer -1 [1,∞)[1, \infty)
nbins integer 20 [1,∞)[1, \infty)
nbins_cats integer 1024 [1,∞)[1, \infty)
nbins_top_level integer 1024 [1,∞)[1, \infty)
ntrees integer 50 [1,∞)[1, \infty)
sample_rate numeric 0.632 [0,1][0, 1]
sample_rate_per_class untyped NULL -
score_each_iteration logical FALSE TRUE, FALSE -
score_tree_interval integer 0 [0,∞)[0, \infty)
seed integer -1 (−∞,∞)(-\infty, \infty)
stopping_metric character AUTO AUTO, logloss, AUC, AUCPR, lift_top_group, misclassification, mean_per_class_error -
stopping_rounds integer 0 [0,∞)[0, \infty)
stopping_tolerance numeric 0.001 [0,∞)[0, \infty)
verbose logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifH2ORandomForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifH2ORandomForest$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifH2ORandomForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Fryda T, LeDell E, Gill N, Aiello S, Fu A, Candel A, Click C, Kraljevic T, Nykodym T, Aboyoun P, Kurka M, Malohlava M, Poirier S, Wong W (2025). h2o: R Interface for the 'H2O' Scalable Machine Learning Platform. R package version 3.46.0.9, https://github.com/h2oai/h2o-3.

See Also

Examples

# Define the Learner
learner = lrn("classif.h2o.randomForest")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification IBk Learner

Description

Instance based algorithm: K-nearest neighbours regression. Calls RWeka::IBk() from RWeka.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.IBk")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
weight character - I, F -
K integer 1 [1,∞)[1, \infty)
E logical FALSE TRUE, FALSE -
W integer 0 [0,∞)[0, \infty)
X logical FALSE TRUE, FALSE -
A character LinearNNSearch BallTree, CoverTree, FilteredNeighbourSearch, KDTree, LinearNNSearch -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

  • weight:

    • original id: I and F

  • Reason for change: original I and F params are interdependent (I can only be TRUE when F is FALSE and vice versa). The easiest way to encode this is to combine I and F into one factor param.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifIBk

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifIBk$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifIBk$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifIBk$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifIBk$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

henrifnk

References

Aha, W D, Kibler, Dennis, Albert, K M (1991). “Instance-based learning algorithms.” Machine learning, 6(1), 37–66.

See Also

Examples

# Define the Learner
learner = lrn("classif.IBk")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Imbalanced Random Forest Src Learner

Description

Imbalanced Random forest for classification between two classes. Calls randomForestSRC::imbalanced.rfsrc() from from randomForestSRC.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.imbalanced_rfsrc")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, randomForestSRC

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
method character rfq rfq, brf, standard -
block.size integer 10 [1,∞)[1, \infty)
fast logical FALSE TRUE, FALSE -
ratio numeric - [0,1][0, 1]
mtry integer - [1,∞)[1, \infty)
mtry.ratio numeric - [0,1][0, 1]
nodesize integer 15 [1,∞)[1, \infty)
nodedepth integer - [1,∞)[1, \infty)
splitrule character gini gini, auc, entropy -
nsplit integer 10 [0,∞)[0, \infty)
importance character FALSE FALSE, TRUE, none, permute, random, anti -
bootstrap character by.root by.root, by.node, none, by.user -
samptype character swor swor, swr -
samp untyped - -
membership logical FALSE TRUE, FALSE -
sampsize untyped - -
sampsize.ratio numeric - [0,1][0, 1]
na.action character na.omit na.omit, na.impute -
nimpute integer 1 [1,∞)[1, \infty)
ntime integer - [1,∞)[1, \infty)
proximity character FALSE FALSE, TRUE, inbag, oob, all -
distance character FALSE FALSE, TRUE, inbag, oob, all -
forest.wt character FALSE FALSE, TRUE, inbag, oob, all -
xvar.wt untyped - -
split.wt untyped - -
forest logical TRUE TRUE, FALSE -
var.used character FALSE FALSE, all.trees -
split.depth character FALSE FALSE, all.trees -
seed integer - (−∞,−1](-\infty, -1]
do.trace logical FALSE TRUE, FALSE -
get.tree untyped - -
outcome character train train, test -
ptn.count integer 0 [0,∞)[0, \infty)
cores integer 1 [1,∞)[1, \infty)
save.memory logical FALSE TRUE, FALSE -
perf.type character - gmean, misclass, brier, none -
case.depth logical FALSE TRUE, FALSE -
marginal.xvar untyped NULL -

Custom mlr3 parameters

  • mtry: This hyperparameter can alternatively be set via the added hyperparameter mtry.ratio as mtry = max(ceiling(mtry.ratio * n_features), 1). Note that mtry and mtry.ratio are mutually exclusive.

  • sampsize: This hyperparameter can alternatively be set via the added hyperparameter sampsize.ratio as sampsize = max(ceiling(sampsize.ratio * n_obs), 1). Note that sampsize and sampsize.ratio are mutually exclusive.

  • cores: This value is set as the option rf.cores during training and is set to 1 by default.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifImbalancedRandomForestSRC

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifImbalancedRandomForestSRC$new()

Method importance()

The importance scores are extracted from the slot importance.

Usage
LearnerClassifImbalancedRandomForestSRC$importance()
Returns

Named numeric().

Method selected_features()

Selected features are extracted from the model slot var.used.

Note: Due to a known issue in randomForestSRC, enabling var.used = "all.trees" causes prediction to fail. Therefore, this setting should be used exclusively for feature selection purposes and not when prediction is required.

Usage
LearnerClassifImbalancedRandomForestSRC$selected_features()
Returns

character().

Method oob_error()

OOB error extracted from the model slot err.rate.

Usage
LearnerClassifImbalancedRandomForestSRC$oob_error()
Returns

numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifImbalancedRandomForestSRC$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

HarutyunyanLiana

References

O’Brien R, Ishwaran H (2019). “A random forests quantile classifier for class imbalanced data.” Pattern Recognition, 90, 232–249. doi:10.1016/j.patcog.2019.01.036.

Chao C, Leo B (2004). “Using Random Forest to Learn Imbalanced Data.” University of California, Berkeley.

See Also

Examples

# Define the Learner
learner = lrn("classif.imbalanced_rfsrc", importance = "TRUE")
print(learner)

# Define a Task
task = tsk("sonar")
# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification J48 Learner

Description

Decision tree algorithm. Calls RWeka::IBk() from RWeka.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.J48")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
U logical FALSE TRUE, FALSE -
O logical FALSE TRUE, FALSE -
C numeric 0.25 [2.22044604925031e−16,1][2.22044604925031e-16, 1]
M integer 2 [1,∞)[1, \infty)
R logical FALSE TRUE, FALSE -
N integer 3 [2,∞)[2, \infty)
B logical FALSE TRUE, FALSE -
S logical FALSE TRUE, FALSE -
L logical FALSE TRUE, FALSE -
A logical FALSE TRUE, FALSE -
J logical FALSE TRUE, FALSE -
Q integer 1 [1,∞)[1, \infty)
doNotMakeSplitPointActualValue logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifJ48

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifJ48$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifJ48$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifJ48$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifJ48$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

henrifnk

References

Quinlan, Ross J (2014). C4. 5: programs for machine learning. Elsevier.

See Also

Examples

# Define the Learner
learner = lrn("classif.J48")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification JRip Learner

Description

Repeated Incremental Pruning to Produce Error Reduction. Calls RWeka::JRip() from RWeka.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.JRip")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
F integer 3 [2,∞)[2, \infty)
N numeric 2 [0,∞)[0, \infty)
O integer 2 [1,∞)[1, \infty)
D logical FALSE TRUE, FALSE -
S integer 1 [1,∞)[1, \infty)
E logical FALSE TRUE, FALSE -
P logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Parameter changes

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifJRip

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifJRip$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifJRip$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifJRip$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifJRip$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

henrifnk

References

Cohen, W W (1995). “Fast effective rule induction.” In Machine learning proceedings 1995, 115–123. Elsevier.

See Also

Examples

# Define the Learner
learner = lrn("classif.JRip")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification K-nearest Neighbour Learner

Description

K-nearest neighbor classification. Calls class::knn() from class.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.knn")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”

  • Feature Types: “numeric”

  • Required Packages: mlr3, class

Parameters

Id Type Default Levels Range
k integer 1 [1,∞)[1, \infty)
l numeric 0 [0,∞)[0, \infty)
prob logical FALSE TRUE, FALSE -
use.all logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifKnn

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifKnn$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifKnn$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Venables WN, Ripley BD (2002). Modern Applied Statistics with S, Fourth edition. Springer, New York. ISBN 0-387-95457-0, https://www.stats.ox.ac.uk/pub/MASS4/.

See Also

Examples

# Define the Learner
learner = lrn("classif.knn")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification KStar Learner

Description

Instance-based classifier which differs from other instance-based learners in that it uses an entropy-based distance function. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.kstar")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
B integer 20 (−∞,∞)(-\infty, \infty)
E logical - TRUE, FALSE -
M character a a, d, m, n -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifKStar

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifKStar$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifKStar$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifKStar$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifKStar$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Cleary JG, Trigg LE (1995). “K*: An Instance-based Learner Using an Entropic Distance Measure.” In 12th International Conference on Machine Learning, 108-114.

See Also

Examples

# Define the Learner
learner = lrn("classif.kstar")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Kernlab Support Vector Machine

Description

Support vector machine for classification. Calls kernlab::ksvm() from kernlab.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.ksvm")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, kernlab

Parameters

Id Type Default Levels Range
scaled logical TRUE TRUE, FALSE -
type character C-svc C-svc, nu-svc, C-bsvc, spoc-svc, kbb-svc -
kernel character rbfdot rbfdot, polydot, vanilladot, laplacedot, besseldot, anovadot -
C numeric 1 (−∞,∞)(-\infty, \infty)
nu numeric 0.2 [0,∞)[0, \infty)
cache integer 40 [1,∞)[1, \infty)
tol numeric 0.001 [0,∞)[0, \infty)
shrinking logical TRUE TRUE, FALSE -
sigma numeric - [0,∞)[0, \infty)
degree integer - [1,∞)[1, \infty)
scale numeric - [0,∞)[0, \infty)
order integer - (−∞,∞)(-\infty, \infty)
offset numeric - (−∞,∞)(-\infty, \infty)
coupler character minpair minpair, pkpd -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifKSVM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifKSVM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifKSVM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

mboecker

References

Karatzoglou, Alexandros, Smola, Alex, Hornik, Kurt, Zeileis, Achim (2004). “kernlab-an S4 package for kernel methods in R.” Journal of statistical software, 11(9), 1–20.

See Also

Examples

# Define the Learner
learner = lrn("classif.ksvm")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

LiblineaR Classification Learner

Description

L2 regularized support vector classification. Calls LiblineaR::LiblineaR() from LiblineaR.

Details

Type of SVC depends on type argument:

  • 0 – L2-regularized logistic regression (primal)

  • 1 - L2-regularized L2-loss support vector classification (dual)

  • 3 - L2-regularized L1-loss support vector classification (dual)

  • 2 – L2-regularized L2-loss support vector classification (primal)

  • 4 – Support vector classification by Crammer and Singer

  • 5 - L1-regularized L2-loss support vector classification

  • 6 - L1-regularized logistic regression

  • 7 - L2-regularized logistic regression (dual)

If number of records > number of features, type = 2 is faster than type = 1 (Hsu et al. 2003).

Note that probabilistic predictions are only available for types 0, 6, and 7. The default epsilon value depends on the type parameter, see LiblineaR::LiblineaR().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.liblinear")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “numeric”

  • Required Packages: mlr3, mlr3extralearners, LiblineaR

Parameters

Id Type Default Levels Range
type integer 0 [0,7][0, 7]
cost numeric 1 [0,∞)[0, \infty)
epsilon numeric - [0,∞)[0, \infty)
bias numeric 1 (−∞,∞)(-\infty, \infty)
cross integer 0 [0,∞)[0, \infty)
verbose logical FALSE TRUE, FALSE -
wi untyped NULL -
findC logical FALSE TRUE, FALSE -
useInitC logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifLiblineaR

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifLiblineaR$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifLiblineaR$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Fan, Rong-En, Chang, Kai-Wei, Hsieh, Cho-Jui, Wang, Xiang-Rui, Lin, Chih-Jen (2008). “LIBLINEAR: A library for large linear classification.” the Journal of machine Learning research, 9, 1871–1874.

See Also

Examples

# Define the Learner
learner = lrn("classif.liblinear")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification LightGBM Learner

Description

Gradient boosting algorithm. Calls lightgbm::lightgbm() from lightgbm. The list of parameters can be found here and in the documentation of lightgbm::lgb.train().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.lightgbm")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, lightgbm

Parameters

Id Type Default Levels Range
objective character - binary, multiclass, multiclassova -
eval untyped - -
verbose integer 1 (−∞,∞)(-\infty, \infty)
record logical TRUE TRUE, FALSE -
eval_freq integer 1 [1,∞)[1, \infty)
callbacks untyped - -
reset_data logical FALSE TRUE, FALSE -
boosting character gbdt gbdt, rf, dart, goss -
linear_tree logical FALSE TRUE, FALSE -
learning_rate numeric 0.1 [0,∞)[0, \infty)
num_leaves integer 31 [1,131072][1, 131072]
tree_learner character serial serial, feature, data, voting -
num_threads integer 0 [0,∞)[0, \infty)
device_type character cpu cpu, gpu -
seed integer - (−∞,∞)(-\infty, \infty)
deterministic logical FALSE TRUE, FALSE -
data_sample_strategy character bagging bagging, goss -
force_col_wise logical FALSE TRUE, FALSE -
force_row_wise logical FALSE TRUE, FALSE -
histogram_pool_size numeric -1 (−∞,∞)(-\infty, \infty)
max_depth integer -1 (−∞,∞)(-\infty, \infty)
min_data_in_leaf integer 20 [0,∞)[0, \infty)
min_sum_hessian_in_leaf numeric 0.001 [0,∞)[0, \infty)
bagging_fraction numeric 1 [0,1][0, 1]
pos_bagging_fraction numeric 1 [0,1][0, 1]
neg_bagging_fraction numeric 1 [0,1][0, 1]
bagging_freq integer 0 [0,∞)[0, \infty)
bagging_seed integer 3 (−∞,∞)(-\infty, \infty)
bagging_by_query logical FALSE TRUE, FALSE -
feature_fraction numeric 1 [0,1][0, 1]
feature_fraction_bynode numeric 1 [0,1][0, 1]
feature_fraction_seed integer 2 (−∞,∞)(-\infty, \infty)
extra_trees logical FALSE TRUE, FALSE -
extra_seed integer 6 (−∞,∞)(-\infty, \infty)
max_delta_step numeric 0 (−∞,∞)(-\infty, \infty)
lambda_l1 numeric 0 [0,∞)[0, \infty)
lambda_l2 numeric 0 [0,∞)[0, \infty)
linear_lambda numeric 0 [0,∞)[0, \infty)
min_gain_to_split numeric 0 [0,∞)[0, \infty)
drop_rate numeric 0.1 [0,1][0, 1]
max_drop integer 50 (−∞,∞)(-\infty, \infty)
skip_drop numeric 0.5 [0,1][0, 1]
xgboost_dart_mode logical FALSE TRUE, FALSE -
uniform_drop logical FALSE TRUE, FALSE -
drop_seed integer 4 (−∞,∞)(-\infty, \infty)
top_rate numeric 0.2 [0,1][0, 1]
other_rate numeric 0.1 [0,1][0, 1]
min_data_per_group integer 100 [1,∞)[1, \infty)
max_cat_threshold integer 32 [1,∞)[1, \infty)
cat_l2 numeric 10 [0,∞)[0, \infty)
cat_smooth numeric 10 [0,∞)[0, \infty)
max_cat_to_onehot integer 4 [1,∞)[1, \infty)
top_k integer 20 [1,∞)[1, \infty)
monotone_constraints untyped NULL -
monotone_constraints_method character basic basic, intermediate, advanced -
monotone_penalty numeric 0 [0,∞)[0, \infty)
feature_contri untyped NULL -
forcedsplits_filename untyped "" -
refit_decay_rate numeric 0.9 [0,1][0, 1]
cegb_tradeoff numeric 1 [0,∞)[0, \infty)
cegb_penalty_split numeric 0 [0,∞)[0, \infty)
cegb_penalty_feature_lazy untyped - -
cegb_penalty_feature_coupled untyped - -
path_smooth numeric 0 [0,∞)[0, \infty)
interaction_constraints untyped - -
use_quantized_grad logical TRUE TRUE, FALSE -
num_grad_quant_bins integer 4 (−∞,∞)(-\infty, \infty)
quant_train_renew_leaf logical FALSE TRUE, FALSE -
stochastic_rounding logical TRUE TRUE, FALSE -
serializable logical TRUE TRUE, FALSE -
max_bin integer 255 [2,∞)[2, \infty)
max_bin_by_feature untyped NULL -
min_data_in_bin integer 3 [1,∞)[1, \infty)
bin_construct_sample_cnt integer 200000 [1,∞)[1, \infty)
data_random_seed integer 1 (−∞,∞)(-\infty, \infty)
is_enable_sparse logical TRUE TRUE, FALSE -
enable_bundle logical TRUE TRUE, FALSE -
use_missing logical TRUE TRUE, FALSE -
zero_as_missing logical FALSE TRUE, FALSE -
feature_pre_filter logical TRUE TRUE, FALSE -
pre_partition logical FALSE TRUE, FALSE -
two_round logical FALSE TRUE, FALSE -
forcedbins_filename untyped "" -
is_unbalance logical FALSE TRUE, FALSE -
scale_pos_weight numeric 1 [0,∞)[0, \infty)
sigmoid numeric 1 [0,∞)[0, \infty)
boost_from_average logical TRUE TRUE, FALSE -
eval_at untyped 1:5 -
multi_error_top_k integer 1 [1,∞)[1, \infty)
auc_mu_weights untyped NULL -
num_machines integer 1 [1,∞)[1, \infty)
local_listen_port integer 12400 [1,∞)[1, \infty)
time_out integer 120 [1,∞)[1, \infty)
machines untyped "" -
gpu_platform_id integer -1 (−∞,∞)(-\infty, \infty)
gpu_device_id integer -1 (−∞,∞)(-\infty, \infty)
gpu_device_id_list untyped NULL -
gpu_use_dp logical FALSE TRUE, FALSE -
num_gpu integer 1 [1,∞)[1, \infty)
start_iteration_predict integer 0 (−∞,∞)(-\infty, \infty)
num_iteration_predict integer -1 (−∞,∞)(-\infty, \infty)
pred_early_stop logical FALSE TRUE, FALSE -
pred_early_stop_freq integer 10 (−∞,∞)(-\infty, \infty)
pred_early_stop_margin numeric 10 (−∞,∞)(-\infty, \infty)
num_iterations integer 100 [1,∞)[1, \infty)
early_stopping_rounds integer - [1,∞)[1, \infty)
early_stopping_min_delta numeric - [0,∞)[0, \infty)
first_metric_only logical FALSE TRUE, FALSE -

Initial parameter values

  • num_threads:

    • Actual default: 0L

    • Initial value: 1L

    • Reason for change: Prevents accidental conflicts with future.

  • verbose:

    • Actual default: 1L

    • Initial value: -1L

    • Reason for change: Prevents accidental conflicts with mlr messaging system.

  • objective:

    • Depends on the task: if binary classification, then this parameter is set to "binary", otherwise "multiclasss" and cannot be changed.

Custom mlr3 parameters

  • num_class: This parameter is automatically inferred for multiclass tasks and does not have to be set.

Early Stopping and Validation

Early stopping can be used to find the optimal number of boosting rounds. Set early_stopping_rounds to an integer value to monitor the performance of the model on the validation set while training. For information on how to configure the validation set, see the Validation section of mlr3::Learner. The internal validation measure can be set the eval parameter which should be a list of mlr3::Measures, functions, or strings for the internal lightgbm measures. If first_metric_only = FALSE (default), the learner stops when any metric fails to improve.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifLightGBM

Active bindings

internal_valid_scores

The last observation of the validation scores for all metrics. Extracted from model$evaluation_log

internal_tuned_values

Returns the early stopped iterations if early_stopping_rounds was set during training.

validate

How to construct the internal validation data. This parameter can be either NULL, a ratio, "test", or "predefined".

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifLightGBM$new()

Method importance()

The importance scores are extracted from lbg.importance.

Usage
LearnerClassifLightGBM$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifLightGBM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

kapsner

References

Ke, Guolin, Meng, Qi, Finley, Thomas, Wang, Taifeng, Chen, Wei, Ma, Weidong, Ye, Qiwei, Liu, Tie-Yan (2017). “Lightgbm: A highly efficient gradient boosting decision tree.” Advances in neural information processing systems, 30.

See Also

Examples

# Define the Learner
learner = lrn("classif.lightgbm")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Logistic Model Trees Learner

Description

Classification tree with logistic regression models at the leaves. Calls RWeka::LMT() from RWeka.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.LMT")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
B logical FALSE TRUE, FALSE -
R logical FALSE TRUE, FALSE -
C logical FALSE TRUE, FALSE -
P logical FALSE TRUE, FALSE -
I integer - [1,∞)[1, \infty)
M integer 15 [1,∞)[1, \infty)
W numeric 0 [0,1][0, 1]
A logical FALSE TRUE, FALSE -
doNotMakeSplitPointActualValue logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifLMT

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifLMT$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifLMT$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifLMT$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifLMT$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

henrifnk

References

Landwehr, Niels, Hall, Mark, Frank, Eibe (2005). “Logistic model trees.” Machine learning, 59(1), 161–205.

See Also

Examples

# Define the Learner
learner = lrn("classif.LMT")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Logistic Regression Learner

Description

Multinomial Logistic Regression model with a ridge estimator. Calls RWeka::Logistic() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.logistic")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
C logical FALSE TRUE, FALSE -
R numeric - (−∞,∞)(-\infty, \infty)
M integer -1 (−∞,∞)(-\infty, \infty)
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifLogistic

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifLogistic$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifLogistic$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifLogistic$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifLogistic$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

le Cessie, S., van Houwelingen, J.C. (1992). “Ridge Estimators in Logistic Regression.” Applied Statistics, 41(1), 191-201.

See Also

Examples

# Define the Learner
learner = lrn("classif.logistic")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Least Squares Support Vector Machine Learner

Description

Least squares support vector machine for classification. Calls kernlab::lssvm() from kernlab. Parameters sigma, degree, scale, offset, order, length, lambda, and normalized are added to make tuning kpar easier. If kpar is provided then these new parameters are ignored. If none are provided then the default "automatic" is used for kpar.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.lssvm")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, kernlab

Parameters

Id Type Default Levels Range
scaled untyped TRUE -
kernel character rbfdot rbfdot, polydot, vanilladot, tanhdot, laplacedot, besseldot, anovadot, splinedot, stringdot -
sigma numeric - (−∞,∞)(-\infty, \infty)
degree numeric - (−∞,∞)(-\infty, \infty)
scale numeric - (−∞,∞)(-\infty, \infty)
offset numeric - (−∞,∞)(-\infty, \infty)
order numeric - (−∞,∞)(-\infty, \infty)
length integer - [0,∞)[0, \infty)
lambda numeric - (−∞,∞)(-\infty, \infty)
normalized logical - TRUE, FALSE -
kpar untyped "automatic" -
tau numeric 0.01 (−∞,∞)(-\infty, \infty)
reduced logical TRUE TRUE, FALSE -
rank integer - (−∞,∞)(-\infty, \infty)
delta integer 40 (−∞,∞)(-\infty, \infty)
tol numeric 1e-04 (−∞,∞)(-\infty, \infty)
fit logical TRUE TRUE, FALSE -
na.action untyped na.omit -
coupler character minpair minpair, pkpd -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifLSSVM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifLSSVM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifLSSVM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Karatzoglou, Alexandros, Smola, Alex, Hornik, Kurt, Zeileis, Achim (2004). “kernlab-an S4 package for kernel methods in R.” Journal of statistical software, 11(9), 1–20.

See Also

Examples

# Define the Learner
learner = lrn("classif.lssvm")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Learning Vector Quantization 1

Description

Learning Vector Quantization 1. Calls class::lvqinit(), class::lvq1(), and class::lvqtest() from class.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.lvq1")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, class

Parameters

Id Type Default Range
alpha numeric 0.03 [0,∞)[0, \infty)
k integer 5 [1,∞)[1, \infty)
niter integer NULL [1,∞)[1, \infty)
prior untyped NULL -
size integer NULL [1,∞)[1, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifLvq1

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifLvq1$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifLvq1$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

The learner does not work on tasks with two target groups and one feature.

Author(s)

awinterstetter

References

Kohonen, Teuvo (1990). “The self-organizing map.” Proceedings of the IEEE, 78(9), 1464–1480. doi:10.1109/5.58325.

Kohonen, Teuvo (1995). Self-Organizing Maps, volume 30 series Springer Series in Information Sciences. Springer, Berlin. ISBN 978-3540967120.

See Also

Examples

# Define the Learner
learner = lrn("classif.lvq1")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Discriminant Analysis Learner

Description

Mixture Discriminant Analysis. Calls mda::mda() from mda.

Initial parameter values

  • keep.fitted: Set to FALSE by default for speed.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.mda")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mda

Parameters

Id Type Default Levels Range
criterion character misclassification misclassification, deviance -
dimension integer - [1,∞)[1, \infty)
eps numeric 2.220446e-16 [0,∞)[0, \infty)
iter integer 5 [1,∞)[1, \infty)
keep.fitted logical TRUE TRUE, FALSE -
method character polyreg polyreg, mars, bruto, gen.ridge -
prior numeric - [0,1][0, 1]
start.method character kmeans kmeans, lvq -
sub.df integer - [1,∞)[1, \infty)
subclasses integer 2 (−∞,∞)(-\infty, \infty)
tot.df integer - [1,∞)[1, \infty)
trace logical FALSE TRUE, FALSE -
tries integer 5 [1,∞)[1, \infty)
weights untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifMda

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifMda$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifMda$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

See Also

Examples

# Define the Learner
learner = lrn("classif.mda")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Linear Discriminant Analysis Learner

Description

Minimum Distance Empirical Bayesian classification, designed for small-sample, high-dimensional data. Empirical Bayes estimator where the eigenvalues of the pooled sample covariance matrix are shrunken towards the identity matrix. Calls sparsediscrim::lda_emp_bayes() from FIXME: sparsediscrim.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.mdeb")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, sparsediscrim

Parameters

Id Type Default
prior untyped NULL

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifMdeb

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifMdeb$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifMdeb$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Srivastava, M., Kubokawa, T. (2007). “Comparison of Discrimination Methods for High Dimensional Data.” Journal of the Japanese Statistical Association, 37(1), 123–134.

See Also

Examples

# Define the Learner
learner = lrn("classif.mdeb")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Multilayer Perceptron Learner

Description

Calls RSNNS::mlp() from RSNNS.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.mlp")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, RSNNS

Parameters

Id Type Default Levels Range
hiddenActFunc untyped "Act_Logistic" -
initFunc untyped "Randomize_Weights" -
initFuncParams untyped - -
inputsTest untyped - -
learnFunc untyped "Std_Backpropagation" -
learnFuncParams untyped - -
linOut logical FALSE TRUE, FALSE -
maxit integer 100 [1,∞)[1, \infty)
pruneFunc untyped - -
pruneFuncParams untyped - -
shufflePatterns logical TRUE TRUE, FALSE -
size untyped 5L -
targetsTest untyped - -
updateFunc untyped "Topological_Order" -
updateFuncParams untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifMLP

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifMLP$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifMLP$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Bergmeir, Christoph, Benitez, M. J (2023). “RSNNS: Neural Networks using the Stuttgart Neural Network Simulator (SNNS).” R package version 0.4-17. doi:10.32614/CRAN.package.RSNNS, https://CRAN.R-project.org/package=RSNNS.

See Also

Examples

# Define the Learner
learner = lrn("classif.mlp")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Model-based Recursive Partitioning Learner

Description

Model-based recursive partitioning algorithm. Calls partykit::mob() from mob.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.mob")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, partykit, sandwich, coin

Parameters

Id Type Default Levels Range
rhs untyped - -
fit untyped - -
offset untyped - -
cluster untyped - -
alpha numeric 0.05 [0,1][0, 1]
bonferroni logical TRUE TRUE, FALSE -
minsize integer - [1,∞)[1, \infty)
minsplit integer - [1,∞)[1, \infty)
minbucket integer - [1,∞)[1, \infty)
maxdepth integer Inf [0,∞)[0, \infty)
mtry integer Inf [0,∞)[0, \infty)
trim numeric 0.1 [0,∞)[0, \infty)
breakties logical FALSE TRUE, FALSE -
parm untyped - -
dfsplit integer - [0,∞)[0, \infty)
prune untyped - -
restart logical TRUE TRUE, FALSE -
verbose logical FALSE TRUE, FALSE -
caseweights logical TRUE TRUE, FALSE -
ytype character vector vector, matrix, data.frame -
xtype character matrix vector, matrix, data.frame -
terminal untyped "object" -
inner untyped "object" -
model logical TRUE TRUE, FALSE -
numsplit character left left, center -
catsplit character binary binary, multiway -
vcov character opg opg, info, sandwich -
ordinal character chisq chisq, max, L2 -
nrep integer 10000 [0,∞)[0, \infty)
applyfun untyped - -
cores integer NULL (−∞,∞)(-\infty, \infty)
additional untyped - -
predict_fun untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifMob

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifMob$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifMob$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sumny

References

Hothorn T, Zeileis A (2015). “partykit: A Modular Toolkit for Recursive Partytioning in R.” Journal of Machine Learning Research, 16(118), 3905-3909. http://jmlr.org/papers/v16/hothorn15a.html.

Hothorn T, Hornik K, Zeileis A (2006). “Unbiased Recursive Partitioning: A Conditional Inference Framework.” Journal of Computational and Graphical Statistics, 15(3), 651–674. doi:10.1198/106186006x133933, https://doi.org/10.1198/106186006x133933.

See Also

Examples

library(mlr3)
logit_ = function(y, x, start = NULL, weights = NULL, offset = NULL, ...) {
  glm(y ~ 1, family = binomial, start = start, ...)
}
learner = LearnerClassifMob$new()
learner$param_set$values$rhs = "."
learner$param_set$values$fit = logit_
learner$param_set$values$additional = list(maxit = 100)
learner$feature_types = c("logical", "integer", "numeric", "factor", "ordered")
learner$properties = c("twoclass", "weights")

predict_fun = function(object, newdata, task, .type) {
  p = unname(predict(object, newdata = newdata, type = "response"))
  levs = task$levels(task$target_names)[[1L]]

  if (.type == "response") {
    ifelse(p < 0.5, levs[1L], levs[2L])
  } else {
    y = matrix(c(1 - p, p), ncol = 2L, nrow = length(p))
    colnames(y) = levs
    y
  }
}
task = tsk("breast_cancer")
learner$param_set$values$predict_fun = predict_fun
ids = partition(task)
learner$train(task, row_ids = ids$train)
learner$predict(task, row_ids = ids$test)

Classification MultilayerPerceptron Learner

Description

Classifier that uses backpropagation to learn a multi-layer perceptron. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

  • G removed:

    • GUI will be opened

  • Reason for change: The parameter is removed because we don't want to launch GUI.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.multilayer_perceptron")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
L numeric 0.3 [0,1][0, 1]
M numeric 0.2 [0,1][0, 1]
N integer 500 [1,∞)[1, \infty)
V numeric 0 [0,100][0, 100]
S integer 0 [0,∞)[0, \infty)
E integer 20 [1,∞)[1, \infty)
A logical FALSE TRUE, FALSE -
B logical FALSE TRUE, FALSE -
H untyped "a" -
C logical FALSE TRUE, FALSE -
I logical FALSE TRUE, FALSE -
R logical FALSE TRUE, FALSE -
D logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifMultilayerPerceptron

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifMultilayerPerceptron$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifMultilayerPerceptron$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifMultilayerPerceptron$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifMultilayerPerceptron$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("classif.multilayer_perceptron")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Multinomial Naive Bayes Learner From Weka

Description

Multinomial Naive Bayes classifier. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.naive_bayes_multinomial")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifNaiveBayesMultinomial

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifNaiveBayesMultinomial$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifNaiveBayesMultinomial$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifNaiveBayesMultinomial$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifNaiveBayesMultinomial$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Mccallum A, Nigam K (1998). “A Comparison of Event Models for Naive Bayes Text Classification.” In AAAI-98 Workshop on 'Learning for Text Categorization'.

See Also

Examples

# Define the Learner
learner = lrn("classif.naive_bayes_multinomial")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Naive Bayes Learner From Weka

Description

Naive Bayes Classifier Using Estimator Classes. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.naive_bayes_weka")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
K logical FALSE TRUE, FALSE -
D logical FALSE TRUE, FALSE -
O logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifNaiveBayesWeka

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifNaiveBayesWeka$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifNaiveBayesWeka$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifNaiveBayesWeka$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifNaiveBayesWeka$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

John GH, Langley P (1995). “Estimating Continuous Distributions in Bayesian Classifiers.” In Eleventh Conference on Uncertainty in Artificial Intelligence, 338-345.

See Also

Examples

# Define the Learner
learner = lrn("classif.naive_bayes_weka")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Neural Network Learner

Description

Calls neuralnet::neuralnet() from neuralnet.

Initial parameter values

  • err.fct is set to "ce".

  • linear.output is set to FALSE to perform classification.

  • act.fct is set to "logistic" for classification.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.neuralnet")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, neuralnet

Parameters

Id Type Default Levels Range
act.fct untyped "logistic" -
algorithm character rprop+ backprop, rprop+, rprop-, sag, slr -
constant.weights untyped - -
err.fct character ce sse, ce -
exclude untyped - -
hidden untyped 1L -
learningrate numeric - (−∞,∞)(-\infty, \infty)
learningrate.factor untyped list(minus = 0.5, plus = 1.2) -
learningrate.limit untyped - -
lifesign character none none, minimal, full -
lifesign.step integer 1000 (−∞,∞)(-\infty, \infty)
likelihood logical FALSE TRUE, FALSE -
linear.output logical FALSE TRUE, FALSE -
rep integer 1 [1,∞)[1, \infty)
startweights untyped - -
stepmax numeric 1e+05 (−∞,∞)(-\infty, \infty)
threshold numeric 0.01 (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifNeuralnet

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifNeuralnet$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifNeuralnet$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

See Also

Examples

# Define the Learner
learner = lrn("classif.neuralnet")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Neural Network Learner

Description

Calls deepnet::nn.train() from deepnet.

Initial parameter values

  • output is set to "softmax" for probabilistic classification.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.nnTrain")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, deepnet

Parameters

Id Type Default Levels Range
activationfun character sigm sigm, linear, tanh -
batchsize integer 100 [1,∞)[1, \infty)
hidden untyped 10L -
hidden_dropout numeric 0 [0,1][0, 1]
initB untyped - -
initW untyped - -
learningrate numeric 0.8 [0,∞)[0, \infty)
learningrate_scale numeric 1 [0,∞)[0, \infty)
max_number_of_layers integer - [1,∞)[1, \infty)
momentum numeric 0.5 [0,∞)[0, \infty)
numepochs integer 3 [1,∞)[1, \infty)
output character - sigm, linear, softmax -
visible_dropout numeric 0 [0,1][0, 1]

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifNNTrain

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifNNTrain$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifNNTrain$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

References

Rong, Xiao (2022). “deepnet: Deep Learning Toolkit in R.” R package version 0.2.1. doi:10.32614/CRAN.package.deepnet, https://CRAN.R-project.org/package=deepnet.

See Also

Examples

# Define the Learner
learner = lrn("classif.nnTrain")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification OneR Learner

Description

One Rule classification algorithm that yields an extremely simple model. Calls RWeka::OneR() from RWeka.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.OneR")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
B integer 6 [1,∞)[1, \infty)
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifOneR

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifOneR$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifOneR$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifOneR$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifOneR$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

henrifnk

References

Holte, C R (1993). “Very simple classification rules perform well on most commonly used datasets.” Machine learning, 11(1), 63–90.

See Also

Examples

# Define the Learner
learner = lrn("classif.OneR")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification PART Learner

Description

Regression partition tree. Calls RWeka::PART() from RWeka.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.PART")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
C numeric 0.25 [2.22044604925031e−16,1][2.22044604925031e-16, 1]
M integer 2 [1,∞)[1, \infty)
R logical FALSE TRUE, FALSE -
N integer 3 [1,∞)[1, \infty)
B logical FALSE TRUE, FALSE -
U logical FALSE TRUE, FALSE -
J logical FALSE TRUE, FALSE -
Q integer 1 [1,∞)[1, \infty)
doNotMakeSplitPointActualValue logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifPART

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifPART$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifPART$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifPART$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifPART$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

henrifnk

References

Frank, Eibe, Witten, H I (1998). “Generating accurate rule sets without global optimization.”

See Also

Examples

# Define the Learner
learner = lrn("classif.PART")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification L1 and L2 Penalized Learner

Description

Penalized (L1 and L2) Logistic Regression. Calls penalized::penalized() from: penalized.

Details

The penalized and unpenalized arguments in the learner are implemented slightly differently than in penalized::penalized(). Here, there is no parameter for penalized but instead it is assumed that every variable is penalized unless stated in the unpenalized parameter.

Initial parameter values

  • trace is set to "FALSE" to disable printing output during model training.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.penalized")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, penalized

Parameters

Id Type Default Levels Range
epsilon numeric 1e-10 [0,∞)[0, \infty)
fusedl logical FALSE TRUE, FALSE -
lambda1 numeric 0 [0,∞)[0, \infty)
lambda2 numeric 0 [0,∞)[0, \infty)
maxiter integer - [1,∞)[1, \infty)
positive untyped FALSE -
standardize logical FALSE TRUE, FALSE -
startbeta untyped - -
startgamma untyped - -
steps untyped 1L -
trace logical TRUE TRUE, FALSE -
unpenalized untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifPenalized

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifPenalized$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifPenalized$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Goeman, J J (2010). “L1 penalized estimation in the Cox proportional hazards model.” Biometrical journal, 52(1), 70–84.

See Also

Examples

# Define the Learner
learner = lrn("classif.penalized")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Partial Least Squares Discriminant Analysis Learner

Description

Partial Least Squares Discriminant Analysis for classification. Calls caret::plsda() from caret.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.plsdaCaret")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, caret, pls

Parameters

Id Type Default Levels Range
ncomp integer 2 [1,∞)[1, \infty)
prior untyped "softmax" -
probMethod character softmax softmax, Bayes -
method character kernelpls kernelpls, widekernelpls, simpls, oscorespls -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifPlsdaCaret

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifPlsdaCaret$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifPlsdaCaret$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Kuhn, Max (2008). “Building Predictive Models in R Using the caret Package.” Journal of Statistical Software, 28(5), 1–26. doi:10.18637/jss.v028.i05.

Mevik, Bjorn-Helge, Wehrens, Ron (2007). “The pls Package: Principal Component and Partial Least Squares Regression in R.” Journal of Statistical Software, 18(2), 1–24. doi:10.18637/jss.v018.i02.

See Also

Examples

# Define the Learner
learner = lrn("classif.plsdaCaret")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Polynomial GLM Learner

Description

Polynomial GLM. Calls polyreg::polyFit() from polyreg.

Initial parameter values

  • deg: We have set this to 2, pretty arbitrarily.

  • noisy: We have set this to FALSE, to get no output on the console.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.polyFit")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”

  • Feature Types: “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, polyreg

Parameters

Id Type Default Levels Range
deg integer - [0,∞)[0, \infty)
maxInteractDeg integer - [0,∞)[0, \infty)
return_xy logical FALSE TRUE, FALSE -
returnPoly logical FALSE TRUE, FALSE -
noisy logical TRUE TRUE, FALSE -
glmMethod character one one, all -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifPolyFit

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifPolyFit$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifPolyFit$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

See Also

Classification Priority Lasso Learner

Description

Patient outcome prediction based on multi-omics data taking practitioners’ preferences into account. Calls prioritylasso::prioritylasso() from prioritylasso.

Initial parameter values

  • family is set to "binomial" and cannot be changed

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.priority_lasso")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, prioritylasso

Parameters

Id Type Default Levels Range
blocks untyped - -
type.measure character - class, auc -
max.coef untyped NULL -
block1.penalization logical TRUE TRUE, FALSE -
lambda.type character lambda.min lambda.min, lambda.1se -
standardize logical TRUE TRUE, FALSE -
nfolds integer 5 [1,∞)[1, \infty)
foldid untyped NULL -
cvoffset logical FALSE TRUE, FALSE -
cvoffsetnfolds integer 10 [1,∞)[1, \infty)
return.x logical TRUE TRUE, FALSE -
handle.missingtestdata character - none, omit.prediction, set.zero, impute.block -
include.allintercepts logical FALSE TRUE, FALSE -
use.blocks untyped "all" -
alignment character lambda lambda, fraction -
alpha numeric 1 [0,1][0, 1]
big numeric 9.9e+35 (−∞,∞)(-\infty, \infty)
devmax numeric 0.999 [0,1][0, 1]
dfmax integer - [0,∞)[0, \infty)
eps numeric 1e-06 [0,1][0, 1]
epsnr numeric 1e-08 [0,1][0, 1]
exclude untyped - -
exmx numeric 250 (−∞,∞)(-\infty, \infty)
fdev numeric 1e-05 [0,1][0, 1]
gamma untyped - -
grouped logical TRUE TRUE, FALSE -
intercept logical TRUE TRUE, FALSE -
keep logical FALSE TRUE, FALSE -
lambda untyped - -
lambda.min.ratio numeric - [0,1][0, 1]
lower.limits untyped -Inf -
maxit integer 100000 [1,∞)[1, \infty)
mnlam integer 5 [1,∞)[1, \infty)
mxit integer 100 [1,∞)[1, \infty)
mxitnr integer 25 [1,∞)[1, \infty)
nlambda integer 100 [1,∞)[1, \infty)
offset untyped NULL -
parallel logical FALSE TRUE, FALSE -
penalty.factor untyped - -
pmax integer - [0,∞)[0, \infty)
pmin numeric 1e-09 [0,1][0, 1]
prec numeric 1e-10 (−∞,∞)(-\infty, \infty)
predict.gamma numeric gamma.1se (−∞,∞)(-\infty, \infty)
relax logical FALSE TRUE, FALSE -
s numeric lambda.1se [0,1][0, 1]
standardize.response logical FALSE TRUE, FALSE -
thresh numeric 1e-07 [0,∞)[0, \infty)
trace.it integer 0 [0,1][0, 1]
type.gaussian character - covariance, naive -
type.logistic character Newton Newton, modified.Newton -
type.multinomial character ungrouped ungrouped, grouped -
upper.limits untyped Inf -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifPriorityLasso

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifPriorityLasso$new()

Method selected_features()

Selected features, i.e. those where the coefficient is non-zero.

Usage
LearnerClassifPriorityLasso$selected_features()
Returns

character().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifPriorityLasso$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

HarutyunyanLiana

References

Simon K, Vindi J, Roman H, Tobias H, Anne-Laure B (2018). “Priority-Lasso: a simple hierarchical approach to the prediction of clinical outcome using multi-omics data.” BMC Bioinformatics, 19. doi:10.1186/s12859-018-2344-6.

See Also

Examples

# Define the Learner and set parameter values
learner = lrn("classif.priority_lasso", type.measure = "auc",
  blocks = list(bp1 = 1:4, bp2 = 5:9, bp3 = 10:28, bp4 = 29:1028))
print(learner)

# Define a Task
task = mlr3::as_task_classif(prioritylasso::pl_data, target = "pl_out")

# Train the learner
learner$train(task)

# print the model
print(learner$model)

Classification Random Forest Learner from Weka

Description

Class for constructing a random forest. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • store_out_of_bag_predictions:

    • original id: store-out-of-bag-predictions

  • output_out_of_bag_complexity_statistics:

    • original id: output-out-of-bag-complexity-statistics

  • num_slots:

    • original id: num-slots

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

  • attribute-importance removed:

    • Compute and output attribute importance (mean impurity decrease method)

  • Reason for change: The parameter is removed because it's unclear how to actually use it.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.random_forest_weka")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
P numeric 100 [0,100][0, 100]
O logical FALSE TRUE, FALSE -
store_out_of_bag_predictions logical FALSE TRUE, FALSE -
output_out_of_bag_complexity_statistics logical FALSE TRUE, FALSE -
print logical FALSE TRUE, FALSE -
I integer 100 [1,∞)[1, \infty)
num_slots integer 1 (−∞,∞)(-\infty, \infty)
K integer 0 (−∞,∞)(-\infty, \infty)
M integer 1 [1,∞)[1, \infty)
V numeric 0.001 (−∞,∞)(-\infty, \infty)
S integer 1 (−∞,∞)(-\infty, \infty)
depth integer 0 [0,∞)[0, \infty)
N integer 0 (−∞,∞)(-\infty, \infty)
U logical FALSE TRUE, FALSE -
B logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifRandomForestWeka

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifRandomForestWeka$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifRandomForestWeka$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifRandomForestWeka$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifRandomForestWeka$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Breiman, Leo (2001). “Random Forests.” Machine Learning, 45(1), 5–32. ISSN 1573-0565, doi:10.1023/A:1010933404324.

See Also

Examples

# Define the Learner
learner = lrn("classif.random_forest_weka")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Random Tree Learner

Description

Tree that considers K randomly chosen attributes at each node. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.random_tree")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
K integer 0 [0,∞)[0, \infty)
M integer 1 [1,∞)[1, \infty)
V numeric 0.001 (−∞,∞)(-\infty, \infty)
S integer 1 (−∞,∞)(-\infty, \infty)
depth integer 0 [0,∞)[0, \infty)
N integer 0 [0,∞)[0, \infty)
U logical FALSE TRUE, FALSE -
B logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifRandomTree

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifRandomTree$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifRandomTree$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifRandomTree$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifRandomTree$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("classif.random_tree")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Random Forest Learner

Description

Random forest for classification. Calls randomForest::randomForest() from randomForest.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.randomForest")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, randomForest

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
mtry integer - [1,∞)[1, \infty)
replace logical TRUE TRUE, FALSE -
classwt untyped NULL -
cutoff untyped - -
strata untyped - -
sampsize untyped - -
nodesize integer 1 [1,∞)[1, \infty)
maxnodes integer - [1,∞)[1, \infty)
importance character FALSE accuracy, gini, none -
localImp logical FALSE TRUE, FALSE -
proximity logical FALSE TRUE, FALSE -
oob.prox logical - TRUE, FALSE -
norm.votes logical TRUE TRUE, FALSE -
do.trace logical FALSE TRUE, FALSE -
keep.forest logical TRUE TRUE, FALSE -
keep.inbag logical FALSE TRUE, FALSE -
predict.all logical FALSE TRUE, FALSE -
nodes logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifRandomForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifRandomForest$new()

Method importance()

The importance scores are extracted from the slot importance. Parameter 'importance' must be set to either "accuracy" or "gini".

Usage
LearnerClassifRandomForest$importance()
Returns

Named numeric().

Method oob_error()

OOB errors are extracted from the model slot err.rate.

Usage
LearnerClassifRandomForest$oob_error()
Returns

numeric(1).

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifRandomForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

pat-s

References

Breiman, Leo (2001). “Random Forests.” Machine Learning, 45(1), 5–32. ISSN 1573-0565, doi:10.1023/A:1010933404324.

See Also

Examples

# Define the Learner
learner = lrn("classif.randomForest", importance = "accuracy")
print(learner)

# Define a Task
task = tsk("sonar")
# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Discriminant Analysis Learner

Description

Regularized Discriminant Analysis. Calls klaR::rda() from: klaR.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.rda")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, klaR

Parameters

Id Type Default Levels Range
lambda numeric - [0,1][0, 1]
gamma numeric - [0,1][0, 1]
crossval logical TRUE TRUE, FALSE -
fold integer 10 [1,∞)[1, \infty)
train.fraction numeric 0.5 [0,1][0, 1]
output logical FALSE TRUE, FALSE -
schedule integer 2 (−∞,∞)(-\infty, \infty)
T.start numeric 0.1 [0,∞)[0, \infty)
halflife numeric 50 [0,∞)[0, \infty)
zero.temp numeric 0.01 [0,∞)[0, \infty)
alpha numeric 2 [1,∞)[1, \infty)
K integer 100 [1,∞)[1, \infty)
trafo logical TRUE TRUE, FALSE -
simAnn logical FALSE TRUE, FALSE -
estimate.error logical TRUE TRUE, FALSE -
max.iter integer 100 (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifRda

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifRda$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifRda$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Friedman JH (1989). “Regularized Discriminant Analysis.” Journal of the American Statistical Association, 84(405), 165–175. ISSN 01621459, 1537274X, http://www.jstor.org/stable/2289860.

See Also

Examples

# Define the Learner
learner = lrn("classif.rda")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Decision Tree Learner

Description

Fast decision tree learner. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.reptree")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
M integer 2 (−∞,∞)(-\infty, \infty)
V numeric 0.001 (−∞,∞)(-\infty, \infty)
N integer 3 (−∞,∞)(-\infty, \infty)
S integer 1 (−∞,∞)(-\infty, \infty)
P logical - TRUE, FALSE -
L integer -1 (−∞,∞)(-\infty, \infty)
I integer 0 (−∞,∞)(-\infty, \infty)
R logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifREPTree

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifREPTree$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifREPTree$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifREPTree$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifREPTree$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("classif.reptree")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Random Ferns Classification Learner

Description

Ensemble machine learning algorithm based on Random Ferns, which are a simplified, faster alternative to Random Forests. Calls rFerns::rFerns() from rFerns.

Initial parameter values

  • importance:

    • Actual default: FALSE

    • Initial value: "simple"

    • Reason for change: The default value of FALSE will resolve to "none", which turns importance calculation off. To enable importance calculation by default, importance is set to "simple".

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.rFerns")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, rFerns

Parameters

Id Type Default Levels Range
consistentSeed untyped NULL -
depth integer 5 [1,16][1, 16]
ferns integer 1000 (−∞,∞)(-\infty, \infty)
importance untyped FALSE -
saveForest logical TRUE TRUE, FALSE -
threads integer 0 (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifRferns

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifRferns$new()

Method importance()

The importance scores are extracted from the model slot importance.

Usage
LearnerClassifRferns$importance()
Returns

Named numeric().

Method oob_error()

OOB error is extracted from the model slot oobErr.

Usage
LearnerClassifRferns$oob_error()
Returns

numeric(1).

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifRferns$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Kursa MB (2014). “rFerns: An Implementation of the Random Ferns Method for General-Purpose Machine Learning.” Journal of Statistical Software, 61(10), 1–13. https://www.jstatsoft.org/v61/i10/.

Ozuysal, Mustafa, Calonder, Michael, Lepetit, Vincent, Fua, Pascal (2010). “Fast Keypoint Recognition Using Random Ferns.” IEEE Transactions on Pattern Analysis and Machine Intelligence, 32(3), 448-461. doi:10.1109/TPAMI.2009.23.

See Also

Examples

# Define the Learner
learner = lrn("classif.rFerns")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Random Forest SRC Learner

Description

Random forest for classification. Calls randomForestSRC::rfsrc() from randomForestSRC.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.rfsrc")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, randomForestSRC

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
mtry integer - [1,∞)[1, \infty)
mtry.ratio numeric - [0,1][0, 1]
nodesize integer 15 [1,∞)[1, \infty)
nodedepth integer - [1,∞)[1, \infty)
splitrule character gini gini, auc, entropy -
nsplit integer 10 [0,∞)[0, \infty)
importance character FALSE FALSE, TRUE, none, permute, random, anti -
block.size integer 10 [1,∞)[1, \infty)
bootstrap character by.root by.root, by.node, none, by.user -
samptype character swor swor, swr -
samp untyped - -
membership logical FALSE TRUE, FALSE -
sampsize untyped - -
sampsize.ratio numeric - [0,1][0, 1]
na.action character na.omit na.omit, na.impute -
nimpute integer 1 [1,∞)[1, \infty)
proximity character FALSE FALSE, TRUE, inbag, oob, all -
distance character FALSE FALSE, TRUE, inbag, oob, all -
forest.wt character FALSE FALSE, TRUE, inbag, oob, all -
xvar.wt untyped - -
split.wt untyped - -
forest logical TRUE TRUE, FALSE -
var.used character FALSE FALSE, all.trees -
split.depth character FALSE FALSE, all.trees, by.tree -
seed integer - (−∞,−1](-\infty, -1]
do.trace logical FALSE TRUE, FALSE -
get.tree untyped - -
outcome character train train, test -
ptn.count integer 0 [0,∞)[0, \infty)
cores integer 1 [1,∞)[1, \infty)
save.memory logical FALSE TRUE, FALSE -
perf.type character - gmean, misclass, brier, none -
case.depth logical FALSE TRUE, FALSE -
marginal.xvar untyped NULL -

Custom mlr3 parameters

  • mtry: This hyperparameter can alternatively be set via the added hyperparameter mtry.ratio as mtry = max(ceiling(mtry.ratio * n_features), 1). Note that mtry and mtry.ratio are mutually exclusive.

  • sampsize: This hyperparameter can alternatively be set via the added hyperparameter sampsize.ratio as sampsize = max(ceiling(sampsize.ratio * n_obs), 1). Note that sampsize and sampsize.ratio are mutually exclusive.

  • cores: This value is set as the option rf.cores during training and is set to 1 by default.

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifRandomForestSRC

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifRandomForestSRC$new()

Method importance()

The importance scores are extracted from the model slot importance, returned for 'all'.

Usage
LearnerClassifRandomForestSRC$importance()
Returns

Named numeric().

Method selected_features()

Selected features are extracted from the model slot var.used.

Note: Due to a known issue in randomForestSRC, enabling var.used = "all.trees" causes prediction to fail. Therefore, this setting should be used exclusively for feature selection purposes and not when prediction is required.

Usage
LearnerClassifRandomForestSRC$selected_features()
Returns

character().

Method oob_error()

OOB error extracted from the model slot err.rate.

Usage
LearnerClassifRandomForestSRC$oob_error()
Returns

numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifRandomForestSRC$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Breiman, Leo (2001). “Random Forests.” Machine Learning, 45(1), 5–32. ISSN 1573-0565, doi:10.1023/A:1010933404324.

See Also

Examples

# Define the Learner
learner = lrn("classif.rfsrc", importance = "TRUE")
print(learner)

# Define a Task
task = tsk("sonar")
# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Rotation Forest Learner

Description

Classification using Rotation Forest algorithm. Training this learner works only for tasks with at least two features. The task must not have utf8 feature names. Calls rotationForest::rotationForest() from rotationForest.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.rotationForest")

Meta Information

Parameters

Id Type Default Range
K integer 3 [1,∞)[1, \infty)
L integer 10 [1,∞)[1, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifRotationForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifRotationForest$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifRotationForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Rodriguez, J.J., Kuncheva, L.I., Alonso, C.J. (2006). “Rotation Forest: A New Classifier Ensemble Method.” IEEE Transactions on Pattern Analysis and Machine Intelligence, 28(10), 1619-1630. doi:10.1109/TPAMI.2006.211.

See Also

Examples

# Define the Learner
learner = lrn("classif.rotationForest")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Stacked Autoencoder Deep Neural Network Learner

Description

Calls deepnet::sae.dnn.train() from deepnet.

Initial parameter values

  • output is set to "softmax" for probabilistic classification.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.saeDNN")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, deepnet

Parameters

Id Type Default Levels Range
hidden untyped 10L -
activationfun character sigm sigm, linear, tanh -
learningrate numeric 0.8 [0,∞)[0, \infty)
momentum numeric 0.5 [0,∞)[0, \infty)
learningrate_scale numeric 1 [0,∞)[0, \infty)
numepochs integer 3 [1,∞)[1, \infty)
batchsize integer 100 [1,∞)[1, \infty)
output character - sigm, linear, softmax -
sae_output character linear sigm, linear, softmax -
hidden_dropout numeric 0 [0,1][0, 1]
visible_dropout numeric 0 [0,1][0, 1]

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifSaeDNN

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifSaeDNN$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifSaeDNN$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Rong, Xiao (2022). “deepnet: Deep Learning Toolkit in R.” R package version 0.2.1. doi:10.32614/CRAN.package.deepnet, https://CRAN.R-project.org/package=deepnet.

See Also

Examples

# Define the Learner
learner = lrn("classif.saeDNN")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Shrinkage Discriminant Analysis

Description

Shrinkage Discriminant Analysis for classification. Calls sda::sda() from sda.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.sda")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, sda

Parameters

Id Type Default Levels Range
lambda numeric - [0,1][0, 1]
lambda.var numeric - [0,1][0, 1]
lambda.freqs numeric - [0,1][0, 1]
diagonal logical FALSE TRUE, FALSE -
verbose logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifSda

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifSda$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifSda$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Ahdesmaeki, Miika, Strimmer, Korbinian (2010). “Feature selection in omics prediction problems using cat scores and false nondiscovery rate control.” The Annals of Applied Statistics, 4(1). ISSN 1932-6157, doi:10.1214/09-aoas277, http://dx.doi.org/10.1214/09-AOAS277.

See Also

Examples

# Define the Learner
learner = lrn("classif.sda")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Linear Discriminant Analysis Learner

Description

Shrinkage-based Diagonal Linear Discriminant Analysis classifier. Type of Naive Bayes classifiers that improves the estimation of the pooled variances by using a shrinkage-based estimator of the pooled covariance matrix. Calls sparsediscrim::lda_shrink_cov() sparsediscrim.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.sdlda")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, sparsediscrim

Parameters

Id Type Default Range
num_alphas integer 101 (−∞,∞)(-\infty, \infty)
prior untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifSdlda

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifSdlda$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifSdlda$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Pang H, Tong T, Zhao H (2009). “Shrinkage-Based Diagonal Discriminant Analysis and Its Applications in High-Dimensional Data.” Biometrics, 65(4), 1021–1029. ISSN 0006341X, 15410420, http://www.jstor.org/stable/20640622.

See Also

Examples

# Define the Learner
learner = lrn("classif.sdlda")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Stochastic Gradient Descent Learner

Description

Stochastic Gradient Descent for learning various linear models. Calls RWeka::make_Weka_classifier() from RWeka.

Initial parameter values

  • F:

    • Has only 2 out of 5 original loss functions: 0 = hinge loss (SVM) and 1 = log loss (logistic regression) with 0 (hinge loss) still being the default

    • Reason for change: this learner should only contain loss functions appropriate for classification tasks

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.sgd")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
F character 0 0, 1 -
L numeric 0.01 (−∞,∞)(-\infty, \infty)
R numeric 1e-04 (−∞,∞)(-\infty, \infty)
E integer 500 (−∞,∞)(-\infty, \infty)
C numeric 0.001 (−∞,∞)(-\infty, \infty)
N logical - TRUE, FALSE -
M logical - TRUE, FALSE -
S integer 1 (−∞,∞)(-\infty, \infty)
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifSGD

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifSGD$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifSGD$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifSGD$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifSGD$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("classif.sgd")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification SimpleLogistic Learner

Description

LogitBoost with simple regression functions as base learners. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.simple_logistic")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
I integer - (−∞,∞)(-\infty, \infty)
S logical FALSE TRUE, FALSE -
P logical FALSE TRUE, FALSE -
M integer - (−∞,∞)(-\infty, \infty)
H integer 50 (−∞,∞)(-\infty, \infty)
W numeric 0 (−∞,∞)(-\infty, \infty)
A logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifSimpleLogistic

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifSimpleLogistic$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifSimpleLogistic$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifSimpleLogistic$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifSimpleLogistic$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Landwehr, Niels, Hall, Mark, Frank, Eibe (2005). “Logistic model trees.” Machine learning, 59(1), 161–205.

Sumner M, Frank E, Hall M (2005). “Speeding up Logistic Model Tree Induction.” In 9th European Conference on Principles and Practice of Knowledge Discovery in Databases, 675-683.

See Also

Examples

# Define the Learner
learner = lrn("classif.simple_logistic")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Support Vector Machine Learner

Description

Support Vector classifier trained with John Platt's sequential minimal optimization algorithm. Calls RWeka::SMO() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • E_poly:

    • original id: E

  • L_poly:

    • original id: L

  • C_poly:

    • original id: C

  • C_logistic:

    • original id: C

  • R_logistic:

    • original id: R

  • M_logistic:

    • original id: M

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.smo")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
no_checks logical FALSE TRUE, FALSE -
C numeric 1 (−∞,∞)(-\infty, \infty)
N character 0 0, 1, 2 -
L numeric 0.001 (−∞,∞)(-\infty, \infty)
P numeric 1e-12 (−∞,∞)(-\infty, \infty)
M logical FALSE TRUE, FALSE -
V integer -1 (−∞,∞)(-\infty, \infty)
W integer 1 (−∞,∞)(-\infty, \infty)
K character PolyKernel NormalizedPolyKernel, PolyKernel, Puk, RBFKernel, StringKernel -
calibrator untyped "weka.classifiers.functions.Logistic" -
E_poly numeric 1 (−∞,∞)(-\infty, \infty)
L_poly logical FALSE TRUE, FALSE -
C_poly integer 25007 (−∞,∞)(-\infty, \infty)
C_logistic logical FALSE TRUE, FALSE -
R_logistic numeric - (−∞,∞)(-\infty, \infty)
M_logistic integer -1 (−∞,∞)(-\infty, \infty)
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifSMO

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifSMO$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifSMO$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifSMO$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifSMO$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Platt J (1998). “Fast Training of Support Vector Machines using Sequential Minimal Optimization.” In Schoelkopf B, Burges C, Smola A (eds.), Advances in Kernel Methods - Support Vector Learning. MIT Press. http://research.microsoft.com/jplatt/smo.html.

Keerthi S, Shevade S, Bhattacharyya C, Murthy K (2001). “Improvements to Platt's SMO Algorithm for SVM Classifier Design.” Neural Computation, 13(3), 637-649.

Hastie T, Tibshirani R (1998). “Classification by Pairwise Coupling.” In Jordan MI, Kearns MJ, Solla SA (eds.), Advances in Neural Information Processing Systems, volume 10.

See Also

Examples

# Define the Learner
learner = lrn("classif.smo")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Sparse Discriminant Analysis

Description

Sparse Linear Discriminant Analysis for classification. Calls sparseLDA::sda() from sparseLDA.

Custom mlr3 parameters

  • Q is set internally to min(n_features, n_classes - 1) when not supplied.

  • stop is not exposed because it depends on the task.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.sparseLDA")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, sparseLDA, MASS, elasticnet

Parameters

Id Type Default Levels Range
lambda numeric 1e-06 [0,∞)[0, \infty)
maxIte integer 100 [0,∞)[0, \infty)
tol numeric 1e-06 [0,∞)[0, \infty)
trace logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifSparseLDA

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifSparseLDA$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifSparseLDA$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Clemmensen, Line, Hastie, Trevor, Witten, Daniela, Ersboll, Bjarne (2011). “Sparse discriminant analysis.” Journal of the American Statistical Association, 106(496), 1519–1531. doi:10.1198/jasa.2011.tm09728.

See Also

Examples

# Define the Learner
learner = lrn("classif.sparseLDA")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Classification Logistic Regression Learner

Description

Logistic regression with a quadratic penalization on the coefficient. Calls stepPlr::plr() from stepPlr.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.stepPlr")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, stepPlr

Parameters

Id Type Default Levels Range
cp character aic aic, bic -
lambda numeric 1e-04 [0,∞)[0, \infty)
offset.coefficients untyped - -
offset.subset untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifStepPlr

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifStepPlr$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifStepPlr$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Park, Young M, Hastie, Trevor (2007). “Penalized logistic regression for detecting gene interactions.” Biostatistics, 9(1), 30-50. ISSN 1465-4644, doi:10.1093/biostatistics/kxm010, https://doi.org/10.1093/biostatistics/kxm010.

See Also

Examples

# Define the Learner
learner = lrn("classif.stepPlr")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

TabPFN Classification Learner

Description

Foundation model for tabular data. Uses reticulate to interface with the tabpfn Python package.

Installation

This learner relies on reticulate to handle Python dependencies. It is not necessary to install any Python package manually in advance or specify a Python environment via reticulate::use_python(), reticulate::use_virtualenv(), reticulate::use_condaenv(), or reticulate::use_miniconda(). By calling ⁠$train()⁠ or ⁠$predict()⁠, the required Python packages (tapfn, torch, etc.) will be installed automatically, if not already. Reticulate will then configure and initialize an ephemeral environment satisfying those requirements, unless an existing environment (e.g., "r-reticulate") in reticulate's Order of Discovery contains all the necessary packages.

You may also manually install tabpfn into a Python environment following the official installation guide and specify the environment via ⁠reticulate::use_*()⁠ before calling ⁠$train()⁠ or ⁠$predict()⁠. Note that the GPU version of PyTorch cannot be loaded by reticulate::use_condaenv() from a conda environment. To use a conda environment, please install the CPU version of PyTorch.

Saving a Learner

In order to save a LearnerClassifTabPFN for later usage, it is necessary to call the ⁠$marshal()⁠ method on the Learner before writing it to disk, as the object will otherwise not be saved correctly. After loading a marshaled LearnerClassifTabPFN into R again, you then need to call ⁠$unmarshal()⁠ to transform it into a useable state.

Initial parameter values

  • n_jobs is initialized to 1 to avoid threading conflicts with future.

Custom mlr3 parameters

  • categorical_feature_indices uses R indexing instead of zero-based Python indexing.

  • device must be a string. If set to "auto", the behavior is the same as original. Otherwise, the string is passed as argument to torch.device() to create a device.

  • inference_precision must be "auto", "autocast", or a torch.dtype string, e.g., "torch.float32", "torch.float64", etc. Non-float dtypes are not supported.

  • inference_config is currently not supported.

  • random_state accepts either an integer or the special value "None" which corresponds to None in Python. Following the original Python implementation, the default random_state is 0.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.tabpfn")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, reticulate

Parameters

Id Type Default Levels Range
n_estimators integer 4 [1,∞)[1, \infty)
categorical_features_indices untyped - -
softmax_temperature numeric 0.9 [0,∞)[0, \infty)
balance_probabilities logical FALSE TRUE, FALSE -
average_before_softmax logical FALSE TRUE, FALSE -
model_path untyped "auto" -
device untyped "auto" -
ignore_pretraining_limits logical FALSE TRUE, FALSE -
inference_precision character auto auto, autocast, torch.float32, torch.float, torch.float64, torch.double, torch.float16, torch.half, torch.bfloat16 -
fit_mode character fit_preprocessors low_memory, fit_preprocessors, fit_with_cache -
memory_saving_mode untyped "auto" -
random_state integer 0 (−∞,∞)(-\infty, \infty)
n_jobs integer - [1,∞)[1, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifTabPFN

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifTabPFN$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifTabPFN$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifTabPFN$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifTabPFN$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

b-zhou

References

Hollmann, Noah, Müller, Samuel, Purucker, Lennart, Krishnakumar, Arjun, Körfer, Max, Hoo, Bin S, Schirrmeister, Tibor R, Hutter, Frank (2025). “Accurate predictions on small data with a tabular foundation model.” Nature. doi:10.1038/s41586-024-08328-6, https://www.nature.com/articles/s41586-024-08328-6.

Hollmann, Noah, Müller, Samuel, Eggensperger, Katharina, Hutter, Frank (2023). “TabPFN: A transformer that solves small tabular classification problems in a second.” In International Conference on Learning Representations 2023.

See Also

Classification Voted Perceptron Learner

Description

Voted Perceptron Algorithm by Freund and Schapire. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("classif.voted_perceptron")

Meta Information

  • Task type: “classif”

  • Predict Types: “response”, “prob”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
I integer 1 [1,∞)[1, \infty)
E numeric 1 (−∞,∞)(-\infty, \infty)
S integer 1 (−∞,∞)(-\infty, \infty)
M integer 10000 (−∞,∞)(-\infty, \infty)
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerClassif -> LearnerClassifVotedPerceptron

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerClassifVotedPerceptron$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerClassifVotedPerceptron$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerClassifVotedPerceptron$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerClassifVotedPerceptron$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Freund Y, Schapire RE (1998). “Large margin classification using the perceptron algorithm.” In 11th Annual Conference on Computational Learning Theory, 209-217.

See Also

Examples

# Define the Learner
learner = lrn("classif.voted_perceptron")
print(learner)

# Define a Task
task = tsk("sonar")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Competing Risks Cox Learner with Likelihood Based Boosting

Description

Fits a competing risks Cox model using likelihood based boosting. Calls CoxBoost::CoxBoost() or CoxBoost::cv.CoxBoost() from package CoxBoost. Predicts cumulative incidence functions (one CIF matrix per cause, rows = observations, columns = times). If times are not provided for prediction, the unique event times (across causes) from the training set are used by default.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("cmprsk.coxboost")

Meta Information

Parameters

Id Type Default Levels Range
cv logical - TRUE, FALSE -
maxstepno integer 100 [0,∞)[0, \infty)
K integer 10 [2,∞)[2, \infty)
type character verweij verweij, naive -
folds untyped NULL -
minstepno integer 50 [0,∞)[0, \infty)
start.penalty numeric - (−∞,∞)(-\infty, \infty)
iter.max integer 10 [1,∞)[1, \infty)
upper.margin numeric 0.05 [0,1][0, 1]
unpen.index untyped - -
standardize logical TRUE TRUE, FALSE -
stepno integer 100 [0,∞)[0, \infty)
penalty numeric - (−∞,∞)(-\infty, \infty)
criterion character pscore pscore, score, hpscore, hscore -
stepsize.factor numeric 1 (−∞,∞)(-\infty, \infty)
sf.scheme character sigmoid sigmoid, linear -
pendistmat untyped - -
connected.index untyped - -
x.is.01 logical FALSE TRUE, FALSE -
return.score logical TRUE TRUE, FALSE -
trace logical FALSE TRUE, FALSE -
cmprsk character - csh, ccsh -
coupled.strata logical TRUE TRUE, FALSE -
at.step untyped - -
times untyped - -

Custom mlr3 parameters

  • cv: If TRUE, uses cross-validation with CoxBoost::cv.CoxBoost() to select the optimal number of boosting steps, then refits the final model using CoxBoost::CoxBoost() with this value. Initialized to FALSE by default.

  • penalty: If set to "optimCoxBoostPenalty", uses CoxBoost::optimCoxBoostPenalty() to determine the optimal penalty value and number of steps (via CoxBoost::cv.CoxBoost()) for the final model.

Initial parameter values

  • cmprsk: Type of competing risks analysis. Initialized to "csh" (cause-specific hazards) which models each cause separately, treating other causes as censored. "ccsh" (combined cause-specific hazards) is another option which combines information across causes for variable selection using a joint criterion.

Super classes

mlr3::Learner -> mlr3cmprsk::LearnerCompRisks -> LearnerCompRisksCoxboost

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerCompRisksCoxboost$new()

Method selected_features()

Returns the set of cause-specific selected features which have non-zero coefficients. Calls the internal coef.CoxBoost() function.

Usage
LearnerCompRisksCoxboost$selected_features(at_step = NULL, cause = 1)
Arguments
at_step

(integer(1))
Which boosting step to get the coefficients for. If no step is given (default), the final boosting step is used.

cause

(integer(1))
Integer value indicating the event of interest.

Returns

(character()) vector of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerCompRisksCoxboost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Binder, Harald, Allignol, Arthur, Schumacher, Martin, Beyersmann, Jan (2009). “Boosting for high-dimensional time-to-event data with competing risks.” Bioinformatics, 25(7), 890–896.

See Also

Examples

# Define the Learner
learner = lrn("cmprsk.coxboost", stepno = 10)
print(learner)

# Define a Task
task = tsk("pbc")
task$col_roles$feature = c("age", "chol", "copper", "platelet", "trig",
 "albumin", "alk.phos", "ast", "bili", "protime")

# Stratification based on event
task$set_col_roles(cols = "status", add_to = "stratum")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$selected_features(cause = 1)) # for cause = 1
print(learner$selected_features(cause = 2)) # for cause = 2

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Random Forest Competing Risks Learner

Description

Random survival forests for competing risks. Calls randomForestSRC::rfsrc() from randomForestSRC.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("cmprsk.rfsrc")

Meta Information

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
mtry integer - [1,∞)[1, \infty)
mtry.ratio numeric - [0,1][0, 1]
nodesize integer 15 [1,∞)[1, \infty)
nodedepth integer - [1,∞)[1, \infty)
splitrule character logrankCR logrankCR, logrank -
nsplit integer 10 [0,∞)[0, \infty)
importance character FALSE FALSE, TRUE, none, anti, permute, random -
block.size integer 10 [1,∞)[1, \infty)
bootstrap character by.root by.root, by.node, none, by.user -
samptype character swor swor, swr -
samp untyped - -
membership logical FALSE TRUE, FALSE -
sampsize untyped - -
sampsize.ratio numeric - [0,1][0, 1]
na.action character na.omit na.omit, na.impute -
nimpute integer 1 [1,∞)[1, \infty)
ntime integer 150 [0,∞)[0, \infty)
cause untyped - -
proximity character FALSE FALSE, TRUE, inbag, oob, all -
distance character FALSE FALSE, TRUE, inbag, oob, all -
forest.wt character FALSE FALSE, TRUE, inbag, oob, all -
xvar.wt untyped - -
split.wt untyped - -
forest logical TRUE TRUE, FALSE -
var.used character FALSE FALSE, all.trees -
split.depth character FALSE FALSE, all.trees, by.tree -
seed integer - (−∞,−1](-\infty, -1]
do.trace logical FALSE TRUE, FALSE -
get.tree untyped - -
outcome character train train, test -
ptn.count integer 0 [0,∞)[0, \infty)
cores integer 1 [1,∞)[1, \infty)
save.memory logical FALSE TRUE, FALSE -
use.uno logical TRUE TRUE, FALSE -
perf.type character - none -
case.depth logical FALSE TRUE, FALSE -
marginal.xvar untyped NULL -

Initial parameter values

  • ntime: Number of time points to coerce the observed event times for use in the estimated cumulative incidence functions during prediction. We changed the default value of 150 to 0, meaning we now use all the unique event times from the train set across all competing causes.

Custom mlr3 parameters

  • mtry: This hyperparameter can alternatively be set via the added hyperparameter mtry.ratio as mtry = max(ceiling(mtry.ratio * n_features), 1). Note that mtry and mtry.ratio are mutually exclusive.

  • sampsize: This hyperparameter can alternatively be set via the added hyperparameter sampsize.ratio as sampsize = max(ceiling(sampsize.ratio * n_obs), 1). Note that sampsize and sampsize.ratio are mutually exclusive.

  • cores: This value is set as the option rf.cores during training and is set to 1 by default.

Super classes

mlr3::Learner -> mlr3cmprsk::LearnerCompRisks -> LearnerCompRisksRandomForestSRC

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerCompRisksRandomForestSRC$new()

Method importance()

The importance scores are extracted from the model slot importance and are cause-specific.

Usage
LearnerCompRisksRandomForestSRC$importance(cause = 1)
Arguments
cause

Integer value indicating the event of interest

Returns

Named numeric().

Method selected_features()

Selected features are extracted from the model slot var.used.

Note: Due to a known issue in randomForestSRC, enabling var.used = "all.trees" causes prediction to fail. Therefore, this setting should be used exclusively for feature selection purposes and not when prediction is required.

Usage
LearnerCompRisksRandomForestSRC$selected_features()
Returns

character().

Method oob_error()

Extracts the out-of-bag (OOB) cumulative incidence function (CIF) error from the model's err.rate slot.

If cause = "mean" (default), the function returns a weighted average of the cause-specific OOB errors, where the weights correspond to the observed proportion of events for each cause in the training data.

Usage
LearnerCompRisksRandomForestSRC$oob_error(cause = "mean")
Arguments
cause

Integer (event type) or "mean" (default). Use a specific event type to retrieve its OOB error, or "mean" to compute the weighted average across causes.

Returns

numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerCompRisksRandomForestSRC$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Ishwaran, H., Gerds, A. T, Kogalur, B. U, Moore, D. R, Gange, J. S, Lau, M. B (2014). “Random survival forests for competing risks.” Biostatistics, 15(4), 757–773. doi:10.1093/BIOSTATISTICS/KXU010, https://doi.org/10.1093/BIOSTATISTICS/KXU010.

See Also

Examples

# Define the Learner
learner = lrn("cmprsk.rfsrc", importance = "TRUE")
print(learner)

# Define a Task
task = tsk("pbc")

# Stratification based on event
task$set_col_roles(cols = "status", add_to = "stratum")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance(cause = 1)) # VIMP for cause = 1
print(learner$importance(cause = 2)) # VIMP for cause = 2
print(learner$oob_error()) # weighted-mean across causes

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Abess Learner

Description

Adaptive best-subset selection for regression. Calls abess::abess() from abess.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.abess")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, abess

Parameters

Id Type Default Levels Range
family character gaussian gaussian, mgaussian, poisson, gamma -
tune.path character sequence sequence, gsection -
tune.type character gic gic, aic, bic, ebic, cv -
normalize integer NULL (−∞,∞)(-\infty, \infty)
support.size untyped NULL -
c.max integer 2 [1,∞)[1, \infty)
gs.range untyped NULL -
lambda numeric 0 [0,∞)[0, \infty)
always.include untyped NULL -
group.index untyped NULL -
init.active.set untyped NULL -
splicing.type integer 2 [1,2][1, 2]
max.splicing.iter integer 20 [1,∞)[1, \infty)
screening.num integer NULL [0,∞)[0, \infty)
important.search integer NULL [0,∞)[0, \infty)
warm.start logical TRUE TRUE, FALSE -
nfolds integer 5 (−∞,∞)(-\infty, \infty)
foldid untyped NULL -
cov.update logical FALSE TRUE, FALSE -
newton character exact exact, approx -
newton.thresh numeric 1e-06 [0,∞)[0, \infty)
max.newton.iter integer NULL [1,∞)[1, \infty)
early.stop logical FALSE TRUE, FALSE -
ic.scale numeric 1 [0,∞)[0, \infty)
num.threads integer 0 [0,∞)[0, \infty)
seed integer 1 (−∞,∞)(-\infty, \infty)

Initial parameter values

  • num.threads: This parameter is initialized to 1 (default is 0) to avoid conflicts with the mlr3 parallelization.

  • family: Depends on the task type, if the parameter family is NULL, it is set to "binomial" for binary classification tasks and to "multinomial" for multiclass classification problems.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrAbess

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrAbess$new()

Method selected_features()

Extract the name of selected features from the model by abess::extract().

Usage
LearnerRegrAbess$selected_features()
Returns

The names of selected features

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrAbess$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bbayukari

See Also

Examples

# Define the Learner
learner = lrn("regr.abess")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Oblique Random Forest Regressor

Description

Accelerated oblique random regression forest. Calls aorsf::orsf() from aorsf. Note that although the learner has the property "missing" and it can in principle deal with missing values, the behaviour has to be configured using the parameter na_action.

Initial parameter values

  • n_thread: This parameter is initialized to 1 (default is 0) to avoid conflicts with the mlr3 parallelization.

  • pred_simplify has to be TRUE, otherwise response is NA in prediction

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.aorsf")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, aorsf

Parameters

Id Type Default Levels Range
attach_data logical TRUE TRUE, FALSE -
epsilon numeric 1e-09 [0,∞)[0, \infty)
importance character anova none, anova, negate, permute -
importance_max_pvalue numeric 0.01 [1e−04,0.9999][1e-04, 0.9999]
leaf_min_events integer 1 [1,∞)[1, \infty)
leaf_min_obs integer 5 [1,∞)[1, \infty)
max_iter integer 20 [1,∞)[1, \infty)
method character glm glm, net, pca, random -
mtry integer NULL [1,∞)[1, \infty)
mtry_ratio numeric - [0,1][0, 1]
n_retry integer 3 [0,∞)[0, \infty)
n_split integer 5 [1,∞)[1, \infty)
n_thread integer - [0,∞)[0, \infty)
n_tree integer 500 [1,∞)[1, \infty)
na_action character fail fail, impute_meanmode -
net_mix numeric 0.5 (−∞,∞)(-\infty, \infty)
oobag logical FALSE TRUE, FALSE -
oobag_eval_every integer NULL [1,∞)[1, \infty)
oobag_fun untyped NULL -
oobag_pred_type character mean none, leaf, mean -
pred_aggregate logical TRUE TRUE, FALSE -
sample_fraction numeric 0.632 [0,1][0, 1]
sample_with_replacement logical TRUE TRUE, FALSE -
scale_x logical FALSE TRUE, FALSE -
split_min_events integer 5 [1,∞)[1, \infty)
split_min_obs integer 10 [1,∞)[1, \infty)
split_min_stat numeric NULL [0,∞)[0, \infty)
split_rule character variance variance -
target_df integer NULL [1,∞)[1, \infty)
tree_seeds integer NULL [1,∞)[1, \infty)
verbose_progress logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrObliqueRandomForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrObliqueRandomForest$new()

Method oob_error()

OOB concordance error extracted from the model slot eval_oobag$stat_values

Usage
LearnerRegrObliqueRandomForest$oob_error()
Returns

numeric().

Method importance()

The importance scores are extracted from the model.

Usage
LearnerRegrObliqueRandomForest$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrObliqueRandomForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

See Also

Examples

# Define the Learner
learner = lrn("regr.aorsf")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression BART (Bayesian Additive Regression Trees) Learner

Description

Bayesian Additive Regression Trees are similar to gradient boosting algorithms. Calls dbarts::bart() from dbarts.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.bart")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, dbarts

Parameters

Id Type Default Levels Range
ntree integer 200 [1,∞)[1, \infty)
sigest untyped NULL -
sigdf integer 3 [1,∞)[1, \infty)
sigquant numeric 0.9 [0,1][0, 1]
k numeric 2 [0,∞)[0, \infty)
power numeric 2 [0,∞)[0, \infty)
base numeric 0.95 [0,1][0, 1]
ndpost integer 1000 [1,∞)[1, \infty)
nskip integer 100 [0,∞)[0, \infty)
printevery integer 100 [0,∞)[0, \infty)
keepevery integer 1 [1,∞)[1, \infty)
keeptrainfits logical TRUE TRUE, FALSE -
usequants logical FALSE TRUE, FALSE -
numcut integer 100 [1,∞)[1, \infty)
printcutoffs integer 0 (−∞,∞)(-\infty, \infty)
verbose logical FALSE TRUE, FALSE -
nthread integer 1 (−∞,∞)(-\infty, \infty)
keeptrees logical FALSE TRUE, FALSE -
keepcall logical TRUE TRUE, FALSE -
sampleronly logical FALSE TRUE, FALSE -
seed integer NA (−∞,∞)(-\infty, \infty)
proposalprobs untyped NULL -
splitprobs untyped NULL -
keepsampler logical - TRUE, FALSE -
n.threads integer - (−∞,∞)(-\infty, \infty)

Custom mlr3 parameters

  • Parameter: offset

    • The parameter is removed, because only dbarts::bart2 allows an offset during training, and therefore the offset parameter in dbarts:::predict.bart is irrelevant for dbarts::dbart.

  • Parameter: nchain, combineChains, combinechains

    • The parameters are removed as parallelization of multiple models is handled by future.

Initial parameter values

  • keeptrees is initialized to TRUE because it is required for prediction.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBart

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBart$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBart$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

ck37

References

Sparapani, Rodney, Spanbauer, Charles, McCulloch, Robert (2021). “Nonparametric machine learning and efficient computation with bayesian additive regression trees: the BART R package.” Journal of Statistical Software, 97, 1–66.

Chipman, A H, George, I E, McCulloch, E R (2010). “BART: Bayesian additive regression trees.” The Annals of Applied Statistics, 4(1), 266–298.

See Also

Examples

# Define the Learner
learner = lrn("regr.bart")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Bayesian CART Learner

Description

Bayesian CART regression model. Calls tgp::bcart() from tgp. For the predicted mean ZZ.km and for the predicted variance ZZ.ks2 are chosen.

Factor features are one-hot encoded with reference encoding before fitting, matching the behavior of the original mlr learner. If factors are present, basemax is set to the number of non-factor features so that tree proposals account for the numeric part of the design.

Initial parameter values

  • verb is initialized to 0 to silence printing.

  • pred.n is initialized to FALSE to skip prediction during training.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.bcart")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, tgp

Parameters

Id Type Default Levels Range
bprior character bflat b0, b0not, bflat, bmle, bmznot, bmzt -
Ds2x logical FALSE TRUE, FALSE -
improv logical FALSE TRUE, FALSE -
itemps untyped NULL -
krige logical TRUE TRUE, FALSE -
m0r1 logical TRUE TRUE, FALSE -
pred.n logical - TRUE, FALSE -
R integer 1 [1,∞)[1, \infty)
trace logical FALSE TRUE, FALSE -
verb integer - [0,4][0, 4]
zcov logical FALSE TRUE, FALSE -
tree untyped c(0.5, 2) -
BTE untyped c(2000L, 7000L, 2L) -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBcart

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBcart$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBcart$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Gramacy RB (2007). “tgp: An R Package for Bayesian Nonstationary, Semiparametric Nonlinear Regression and Design by Treed Gaussian Process Models.” Journal of Statistical Software, 19(9), 1–46. doi:10.18637/jss.v019.i09.

Gramacy RB, Taddy M (2010). “Categorical Inputs, Sensitivity Analysis, Optimization and Importance Tempering with tgp Version 2, an R Package for Treed Gaussian Process Models.” Journal of Statistical Software, 33(6), 1–48. doi:10.18637/jss.v033.i06.

See Also

Examples

# Define the Learner
learner = lrn("regr.bcart")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Bayesian Gaussian Process Learner

Description

Bayesian Gaussian process regression. Calls tgp::bgp() from tgp. For the predicted mean ZZ.km and for the predicted variance ZZ.ks2 are chosen.

Initial parameter values

  • verb is initialized to 0 to silence printing.

  • pred.n is initialized to FALSE to skip prediction during training.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.bgp")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, tgp

Parameters

Id Type Default Levels Range
meanfn character linear constant, linear -
bprior character bflat b0, b0not, bflat, bmle, bmznot, bmzt -
corr character expsep exp, expsep, matern, sim -
BTE untyped c(1000L, 4000L, 2L) -
R integer 1 [1,∞)[1, \infty)
m0r1 logical TRUE TRUE, FALSE -
itemps untyped NULL -
pred.n logical - TRUE, FALSE -
krige logical TRUE TRUE, FALSE -
zcov logical FALSE TRUE, FALSE -
Ds2x logical FALSE TRUE, FALSE -
improv logical FALSE TRUE, FALSE -
nu numeric 1.5 (−∞,∞)(-\infty, \infty)
trace logical FALSE TRUE, FALSE -
verb integer - [0,4][0, 4]
MAP logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBgp

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBgp$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBgp$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Gramacy RB (2007). “tgp: An R Package for Bayesian Nonstationary, Semiparametric Nonlinear Regression and Design by Treed Gaussian Process Models.” Journal of Statistical Software, 19(9), 1–46. doi:10.18637/jss.v019.i09.

Gramacy RB, Taddy M (2010). “Categorical Inputs, Sensitivity Analysis, Optimization and Importance Tempering with tgp Version 2, an R Package for Treed Gaussian Process Models.” Journal of Statistical Software, 33(6), 1–48. doi:10.18637/jss.v033.i06.

See Also

Examples

# Define the Learner
learner = lrn("regr.bgp")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Bayesian Gaussian Process LLM Learner

Description

Bayesian Gaussian process regression with jumps to the limiting linear model. Calls tgp::bgpllm() from tgp. For the predicted mean ZZ.km and for the predicted variance ZZ.ks2 are chosen.

Initial parameter values

  • verb is initialized to 0 to silence printing.

  • pred.n is initialized to FALSE to skip prediction during training.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.bgpllm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, tgp

Parameters

Id Type Default Levels Range
meanfn character linear constant, linear -
bprior character bflat b0, b0not, bflat, bmle, bmznot, bmzt -
corr character expsep exp, expsep, matern, sim -
gamma untyped c(10, 0.2, 0.7) -
BTE untyped c(2000L, 4000L, 2L) -
R integer 1 [1,∞)[1, \infty)
m0r1 logical TRUE TRUE, FALSE -
itemps untyped NULL -
pred.n logical - TRUE, FALSE -
krige logical TRUE TRUE, FALSE -
zcov logical FALSE TRUE, FALSE -
Ds2x logical FALSE TRUE, FALSE -
improv logical FALSE TRUE, FALSE -
nu numeric 1.5 (−∞,∞)(-\infty, \infty)
trace logical FALSE TRUE, FALSE -
verb integer - [0,4][0, 4]
MAP logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBgpllm

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBgpllm$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBgpllm$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Gramacy RB (2007). “tgp: An R Package for Bayesian Nonstationary, Semiparametric Nonlinear Regression and Design by Treed Gaussian Process Models.” Journal of Statistical Software, 19(9), 1–46. doi:10.18637/jss.v019.i09.

Gramacy RB, Taddy M (2010). “Categorical Inputs, Sensitivity Analysis, Optimization and Importance Tempering with tgp Version 2, an R Package for Treed Gaussian Process Models.” Journal of Statistical Software, 33(6), 1–48. doi:10.18637/jss.v033.i06.

See Also

Examples

# Define the Learner
learner = lrn("regr.bgpllm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Bayesian Linear Model Learner

Description

Bayesian linear regression model. Calls tgp::blm() from tgp. For the predicted mean ZZ.km and for the predicted variance ZZ.ks2 are chosen.

Initial parameter values

  • verb is initialized to 0 to silence printing.

  • pred.n is initialized to FALSE to skip prediction during training.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.blm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, tgp

Parameters

Id Type Default Levels Range
bprior character bflat b0, b0not, bflat, bmle, bmznot, bmzt -
Ds2x logical FALSE TRUE, FALSE -
improv logical FALSE TRUE, FALSE -
itemps untyped NULL -
krige logical TRUE TRUE, FALSE -
m0r1 logical TRUE TRUE, FALSE -
MAP logical TRUE TRUE, FALSE -
meanfn character linear constant, linear -
pred.n logical - TRUE, FALSE -
R integer 1 [1,∞)[1, \infty)
trace logical FALSE TRUE, FALSE -
verb integer - [0,4][0, 4]
zcov logical FALSE TRUE, FALSE -
BTE untyped c(1000L, 4000L, 3L) -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBlm

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBlm$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBlm$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Gramacy RB (2007). “tgp: An R Package for Bayesian Nonstationary, Semiparametric Nonlinear Regression and Design by Treed Gaussian Process Models.” Journal of Statistical Software, 19(9), 1–46. doi:10.18637/jss.v019.i09.

Gramacy RB, Taddy M (2010). “Categorical Inputs, Sensitivity Analysis, Optimization and Importance Tempering with tgp Version 2, an R Package for Treed Gaussian Process Models.” Journal of Statistical Software, 33(6), 1–48. doi:10.18637/jss.v033.i06.

See Also

Examples

# Define the Learner
learner = lrn("regr.blm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

BlockForest Regression Learner

Description

Random forests for blocks of clinical and omics covariate data. Calls blockForest::blockfor() from package blockForest. The training model includes only the ⁠$forest⁠ slot, excluding the paramvalues and the biased_oob_error_donotuse.

In this learner, only the trained forest object (⁠$forest⁠) is retained. The optimized block-specific tuning parameters (paramvalues) and the biased OOB error estimate (biased_oob_error_donotuse) are discarded, as they are either not needed for downstream use or not reliable for performance estimation.

Initial parameter values

  • num.threads is initialized to 1 to avoid conflicts with parallelization via future.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.blockforest")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, blockForest

Parameters

Id Type Default Levels Range
blocks untyped - -
block.method character BlockForest BlockForest, RandomBlock, BlockVarSel, VarProb, SplitWeights -
num.trees integer 2000 [1,∞)[1, \infty)
mtry untyped NULL -
nsets integer 300 [1,∞)[1, \infty)
num.trees.pre integer 1500 [1,∞)[1, \infty)
splitrule character extratrees extratrees, variance, maxstat -
always.select.block integer 0 [0,1][0, 1]
importance character - none, impurity, impurity_corrected, permutation -
num.threads integer - [1,∞)[1, \infty)
seed integer NULL (−∞,∞)(-\infty, \infty)
verbose logical TRUE TRUE, FALSE -
se.method character infjack jack, infjack -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBlockForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBlockForest$new()

Method importance()

The importance scores are extracted from the model slot variable.importance.

Usage
LearnerRegrBlockForest$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBlockForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Hornung, R., Wright, N. M (2019). “Block Forests: Random forests for blocks of clinical and omics covariate data.” BMC Bioinformatics, 20(1), 1–17. doi:10.1186/s12859-019-2942-y, https://doi.org/10.1186/s12859-019-2942-y.

See Also

Examples

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# check task's features
task$feature_names

# partition features to 2 blocks
blocks = list(bl1 = 1:3, bl2 = 4:10)

# define learner
learner = lrn("regr.blockforest", blocks = blocks,
              importance = "permutation", nsets = 10,
              num.trees = 50, num.trees.pre = 10, splitrule = "variance")

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# feature importance
learner$importance()

# Make predictions for the test observations
pred = learner$predict(task, row_ids = ids$test)
pred

# Score the predictions
pred$score()

BoTorch Fully Bayesian SingleTaskGP Regression Learner

Description

Fully Bayesian Gaussian Process via botorch, using the SaasFullyBayesianSingleTaskGP model. Unlike standard GP models that use MAP estimation for hyperparameters, this model uses MCMC (NUTS) to learn full posterior distributions over kernel hyperparameters. See here for more details. Uses reticulate to interface with Python.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.botorch_fullybayesian")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, reticulate

Parameters

Id Type Default Levels Range
device character cpu cpu, cuda -
input_transform character - normalize, standardize, log10, warp, none -
outcome_transform character - standardize, none -
warmup_steps integer - [1,∞)[1, \infty)
num_samples integer - [1,∞)[1, \infty)
thinning integer - [1,∞)[1, \infty)
max_tree_depth integer - [1,∞)[1, \infty)
disable_progbar logical - TRUE, FALSE -
jit_compile logical - TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBotorchFullyBayesian

Active bindings

marshaled

(logical(1)) Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBotorchFullyBayesian$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrBotorchFullyBayesian$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrBotorchFullyBayesian$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBotorchFullyBayesian$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

Marc Becker

BoTorch MixedSingleTaskGP Regression Learner

Description

Gaussian Process via botorch and gpytorch, using the MixedSingleTaskGP. See here for more details. Uses reticulate to interface with Python.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.botorch_mixedsingletaskgp")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, reticulate

Parameters

Id Type Default Levels
device character cpu cpu, cuda
input_transform character - normalize, standardize, log10, warp, none
outcome_transform character - standardize, log, none

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBotorchMixedSingleTaskGP

Active bindings

marshaled

(logical(1)) Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBotorchMixedSingleTaskGP$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrBotorchMixedSingleTaskGP$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrBotorchMixedSingleTaskGP$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBotorchMixedSingleTaskGP$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

Marc Becker

BoTorch SingleTaskGP Regression Learner

Description

Gaussian Process via botorch and gpytorch, using the SingleTaskGP model. See here for more details. Uses reticulate to interface with Python.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.botorch_singletaskgp")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, reticulate

Parameters

Id Type Default Levels
device character cpu cpu, cuda
kernel character - matern_2.5, matern_1.5, matern_0.5, rbf, linear, polynomial, periodic, cosine, rq, piecewise_polynomial, [...]
input_transform character - normalize, standardize, log10, warp, none
outcome_transform character - standardize, log, none

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBotorchSingleTaskGP

Active bindings

marshaled

(logical(1)) Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBotorchSingleTaskGP$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrBotorchSingleTaskGP$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrBotorchSingleTaskGP$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBotorchSingleTaskGP$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

Marc Becker

Regression Neural Network Learner

Description

Bayesian regularization for feed-forward neural networks. Calls brnn::brnn() from brnn.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.brnn")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, brnn

Parameters

Id Type Default Levels Range
change numeric 0.001 (−∞,∞)(-\infty, \infty)
cores integer 1 [1,∞)[1, \infty)
epochs integer 1000 [1,∞)[1, \infty)
min_grad numeric 1e-10 (−∞,∞)(-\infty, \infty)
Monte_Carlo logical FALSE TRUE, FALSE -
mu numeric 0.005 (−∞,∞)(-\infty, \infty)
mu_dec numeric 0.1 (−∞,∞)(-\infty, \infty)
mu_inc numeric 10 (−∞,∞)(-\infty, \infty)
mu_max numeric 1e+10 (−∞,∞)(-\infty, \infty)
neurons integer 2 [1,∞)[1, \infty)
normalize logical TRUE TRUE, FALSE -
samples integer 40 [1,∞)[1, \infty)
tol numeric 1e-06 (−∞,∞)(-\infty, \infty)
verbose logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBrnn

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBrnn$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBrnn$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

See Also

Examples

# Define the Learner
learner = lrn("regr.brnn")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Gradient Boosting Learner

Description

Regression gradient boosting learner. Calls bst::bst() from bst.

Initial parameter values

  • Learner = "ls": Default base learner type

  • xval = 0: No cross-validation

  • maxdepth = 1: Maximum tree depth

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.bst")

Meta Information

Parameters

Id Type Default Levels Range
center logical FALSE TRUE, FALSE -
coefir untyped NULL -
cost numeric 0.5 [0,1][0, 1]
cp numeric 0.01 [0,1][0, 1]
df integer 4 [1,∞)[1, \infty)
family character gaussian gaussian, laplace, huber, rhuberDC, thingeDC, tbinomDC, binomdDC -
f.init untyped NULL -
fk untyped NULL -
intercept logical TRUE TRUE, FALSE -
iter integer 1 [1,∞)[1, \infty)
Learner character ls ls, sm, tree -
maxdepth integer 1 [1,30][1, 30]
maxsurrogate integer 5 [0,∞)[0, \infty)
minbucket integer - [1,∞)[1, \infty)
minsplit integer 20 [1,∞)[1, \infty)
mstop integer 50 [1,∞)[1, \infty)
numsample integer 50 [1,∞)[1, \infty)
nu numeric 0.1 [0,1][0, 1]
q numeric - [0,1][0, 1]
qh numeric - [0,1][0, 1]
s numeric - [0,∞)[0, \infty)
sh numeric - [0,∞)[0, \infty)
start logical FALSE TRUE, FALSE -
surrogatestyle integer 0 [0,1][0, 1]
threshold character adaptive adaptive, fixed -
trace logical FALSE TRUE, FALSE -
trun logical FALSE TRUE, FALSE -
twinboost logical FALSE TRUE, FALSE -
twintype integer 1 [1,2][1, 2]
xselect.init untyped NULL -
xval integer 10 [0,∞)[0, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBst

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBst$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBst$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

See Also

Examples

# Define the Learner
learner = lrn("regr.bst")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Bayesian Treed Gaussian Process Learner

Description

Bayesian treed Gaussian process regression model. Calls tgp::btgp() from tgp. For the predicted mean ZZ.km and for the predicted variance ZZ.ks2 are chosen.

Factor features are one-hot encoded with reference encoding before fitting. If factors are present, basemax is set to the number of non-factor features so that tree proposals account for the numeric part of the design.

Initial parameter values

  • verb is initialized to 0 to silence printing.

  • pred.n is initialized to FALSE to skip prediction during training.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.btgp")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, tgp

Parameters

Id Type Default Levels Range
meanfn character linear constant, linear -
bprior character bflat b0, b0not, bflat, bmle, bmznot, bmzt -
corr character expsep exp, expsep, matern, sim -
tree untyped c(0.5, 2) -
BTE untyped c(2000L, 7000L, 2L) -
R integer 1 [1,∞)[1, \infty)
m0r1 logical TRUE TRUE, FALSE -
linburn logical FALSE TRUE, FALSE -
itemps untyped NULL -
pred.n logical - TRUE, FALSE -
krige logical TRUE TRUE, FALSE -
zcov logical FALSE TRUE, FALSE -
Ds2x logical FALSE TRUE, FALSE -
improv logical FALSE TRUE, FALSE -
nu numeric 1.5 (−∞,∞)(-\infty, \infty)
trace logical FALSE TRUE, FALSE -
verb integer - [0,4][0, 4]
MAP logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBtgp

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBtgp$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBtgp$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Gramacy RB (2007). “tgp: An R Package for Bayesian Nonstationary, Semiparametric Nonlinear Regression and Design by Treed Gaussian Process Models.” Journal of Statistical Software, 19(9), 1–46. doi:10.18637/jss.v019.i09.

Gramacy RB, Taddy M (2010). “Categorical Inputs, Sensitivity Analysis, Optimization and Importance Tempering with tgp Version 2, an R Package for Treed Gaussian Process Models.” Journal of Statistical Software, 33(6), 1–48. doi:10.18637/jss.v033.i06.

See Also

Examples

# Define the Learner
learner = lrn("regr.btgp")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Bayesian Treed Gaussian Process LLM Learner

Description

Bayesian treed Gaussian process with jumps to the limiting linear model. Calls tgp::btgpllm() from tgp. For the predicted mean ZZ.km and for the predicted variance ZZ.ks2 are chosen.

Factor features are one-hot encoded with reference encoding before fitting. If factors are present, basemax is set to the number of non-factor features so that tree proposals account for the numeric part of the design.

Initial parameter values

  • verb is initialized to 0 to silence printing.

  • pred.n is initialized to FALSE to skip prediction during training.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.btgpllm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, tgp

Parameters

Id Type Default Levels Range
bprior character bflat b0, b0not, bflat, bmle, bmznot, bmzt -
corr character expsep exp, expsep, matern, sim -
Ds2x logical FALSE TRUE, FALSE -
improv logical FALSE TRUE, FALSE -
itemps untyped NULL -
krige logical TRUE TRUE, FALSE -
linburn logical FALSE TRUE, FALSE -
m0r1 logical TRUE TRUE, FALSE -
MAP logical TRUE TRUE, FALSE -
meanfn character linear constant, linear -
nu numeric 1.5 (−∞,∞)(-\infty, \infty)
pred.n logical - TRUE, FALSE -
R integer 1 [1,∞)[1, \infty)
trace logical FALSE TRUE, FALSE -
verb integer - [0,4][0, 4]
zcov logical FALSE TRUE, FALSE -
tree untyped c(0.5, 2) -
gamma untyped c(10, 0.2, 0.7) -
BTE untyped c(2000L, 7000L, 2L) -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBtgpllm

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBtgpllm$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBtgpllm$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Gramacy RB (2007). “tgp: An R Package for Bayesian Nonstationary, Semiparametric Nonlinear Regression and Design by Treed Gaussian Process Models.” Journal of Statistical Software, 19(9), 1–46. doi:10.18637/jss.v019.i09.

See Also

Examples

# Define the Learner
learner = lrn("regr.btgpllm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Bayesian Treed Linear Model Learner

Description

Bayesian treed linear model regression. Calls tgp::btlm() from tgp. For the predicted mean ZZ.km and for the predicted variance ZZ.ks2 are chosen.

Factor features are one-hot encoded with reference encoding before fitting. If factors are present, basemax is set to the number of non-factor features so that tree proposals account for the numeric part of the design.

Initial parameter values

  • verb is initialized to 0 to silence printing.

  • pred.n is initialized to FALSE to skip prediction during training.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.btlm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, tgp

Parameters

Id Type Default Levels Range
meanfn character linear constant, linear -
bprior character bflat b0, b0not, bflat, bmle, bmznot, bmzt -
tree untyped c(0.5, 2) -
BTE untyped c(2000L, 7000L, 2L) -
R integer 1 [1,∞)[1, \infty)
m0r1 logical TRUE TRUE, FALSE -
itemps untyped NULL -
pred.n logical - TRUE, FALSE -
krige logical TRUE TRUE, FALSE -
zcov logical FALSE TRUE, FALSE -
Ds2x logical FALSE TRUE, FALSE -
improv logical FALSE TRUE, FALSE -
trace logical FALSE TRUE, FALSE -
verb integer - [0,4][0, 4]
MAP logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrBtlm

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrBtlm$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrBtlm$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Gramacy RB (2007). “tgp: An R Package for Bayesian Nonstationary, Semiparametric Nonlinear Regression and Design by Treed Gaussian Process Models.” Journal of Statistical Software, 19(9), 1–46. doi:10.18637/jss.v019.i09.

Gramacy RB, Taddy M (2010). “Categorical Inputs, Sensitivity Analysis, Optimization and Importance Tempering with tgp Version 2, an R Package for Treed Gaussian Process Models.” Journal of Statistical Software, 33(6), 1–48. doi:10.18637/jss.v033.i06.

See Also

Examples

# Define the Learner
learner = lrn("regr.btlm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Gradient Boosted Decision Trees Regression Learner

Description

Gradient boosting algorithm that also supports categorical data. Calls catboost::catboost.train() from package 'catboost'.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.catboost")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, catboost

Parameters

Id Type Default Levels Range
loss_function character RMSE MAE, MAPE, Poisson, Quantile, RMSE, LogLinQuantile, Lq, Huber, Expectile, Tweedie -
learning_rate numeric 0.03 [0.001,1][0.001, 1]
random_seed integer 0 [0,∞)[0, \infty)
l2_leaf_reg numeric 3 [0,∞)[0, \infty)
bootstrap_type character - Bayesian, Bernoulli, MVS, Poisson, No -
bagging_temperature numeric 1 [0,∞)[0, \infty)
subsample numeric - [0,1][0, 1]
sampling_frequency character PerTreeLevel PerTree, PerTreeLevel -
sampling_unit character Object Object, Group -
mvs_reg numeric - [0,∞)[0, \infty)
random_strength numeric 1 [0,∞)[0, \infty)
depth integer 6 [1,16][1, 16]
grow_policy character SymmetricTree SymmetricTree, Depthwise, Lossguide -
min_data_in_leaf integer 1 [1,∞)[1, \infty)
max_leaves integer 31 [1,∞)[1, \infty)
has_time logical FALSE TRUE, FALSE -
rsm numeric 1 [0.001,1][0.001, 1]
nan_mode character Min Min, Max -
fold_permutation_block integer - [1,256][1, 256]
leaf_estimation_method character - Newton, Gradient, Exact -
leaf_estimation_iterations integer - [1,∞)[1, \infty)
leaf_estimation_backtracking character AnyImprovement No, AnyImprovement, Armijo -
fold_len_multiplier numeric 2 [1.001,∞)[1.001, \infty)
approx_on_full_history logical TRUE TRUE, FALSE -
boosting_type character - Ordered, Plain -
boost_from_average logical - TRUE, FALSE -
langevin logical FALSE TRUE, FALSE -
diffusion_temperature numeric 10000 [0,∞)[0, \infty)
score_function character Cosine Cosine, L2, NewtonCosine, NewtonL2 -
monotone_constraints untyped - -
feature_weights untyped - -
first_feature_use_penalties untyped - -
penalties_coefficient numeric 1 [0,∞)[0, \infty)
per_object_feature_penalties untyped - -
model_shrink_rate numeric - (−∞,∞)(-\infty, \infty)
model_shrink_mode character - Constant, Decreasing -
target_border numeric - (−∞,∞)(-\infty, \infty)
border_count integer - [1,65535][1, 65535]
feature_border_type character GreedyLogSum Median, Uniform, UniformAndQuantiles, MaxLogSum, MinEntropy, GreedyLogSum -
per_float_feature_quantization untyped - -
thread_count integer 1 [−1,∞)[-1, \infty)
task_type character CPU CPU, GPU -
devices untyped - -
logging_level character Silent Silent, Verbose, Info, Debug -
metric_period integer 1 [1,∞)[1, \infty)
train_dir untyped "catboost_info" -
model_size_reg numeric 0.5 [0,1][0, 1]
allow_writing_files logical FALSE TRUE, FALSE -
save_snapshot logical FALSE TRUE, FALSE -
snapshot_file untyped - -
snapshot_interval integer 600 [1,∞)[1, \infty)
simple_ctr untyped - -
combinations_ctr untyped - -
ctr_target_border_count integer - [1,255][1, 255]
counter_calc_method character Full SkipTest, Full -
max_ctr_complexity integer - [1,∞)[1, \infty)
ctr_leaf_count_limit integer - [1,∞)[1, \infty)
store_all_simple_ctr logical FALSE TRUE, FALSE -
final_ctr_computation_mode character Default Default, Skip -
verbose logical FALSE TRUE, FALSE -
ntree_start integer 0 [0,∞)[0, \infty)
ntree_end integer 0 [0,∞)[0, \infty)
early_stopping_rounds integer - [1,∞)[1, \infty)
eval_metric untyped - -
use_best_model logical - TRUE, FALSE -
iterations integer 1000 [1,∞)[1, \infty)

Installation

See https://catboost.ai/en/docs/concepts/r-installation.

Initial parameter values

  • logging_level:

    • Actual default: "Verbose"

    • Adjusted default: "Silent"

    • Reason for change: consistent with other mlr3 learners

  • thread_count:

    • Actual default: -1

    • Adjusted default: 1

    • Reason for change: consistent with other mlr3 learners

  • allow_writing_files:

    • Actual default: TRUE

    • Adjusted default: FALSE

    • Reason for change: consistent with other mlr3 learners

  • save_snapshot:

    • Actual default: TRUE

    • Adjusted default: FALSE

    • Reason for change: consistent with other mlr3 learners

Early stopping

Early stopping can be used to find the optimal number of boosting rounds. Set early_stopping_rounds to an integer value to monitor the performance of the model on the validation set while training. For information on how to configure the validation set, see the Validation section of mlr3::Learner.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrCatboost

Active bindings

internal_valid_scores

The last observation of the validation scores for all metrics. Extracted from model$evaluation_log

internal_tuned_values

Returns the early stopped iterations if early_stopping_rounds was set during training.

validate

How to construct the internal validation data. This parameter can be either NULL, a ratio, "test", or "predefined".

Methods

Public methods

Inherited methods

Method new()

Create a LearnerRegrCatboost object.

Usage
LearnerRegrCatboost$new()

Method importance()

The importance scores are calculated using catboost::catboost.get_feature_importance(), setting type = "FeatureImportance", returned for 'all'.

Usage
LearnerRegrCatboost$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrCatboost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sumny

References

Dorogush, Veronika A, Ershov, Vasily, Gulin, Andrey (2018). “CatBoost: gradient boosting with categorical features support.” arXiv preprint arXiv:1810.11363.

See Also

Examples

# Define the Learner
learner = lrn("regr.catboost")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Conditional Random Forest Learner

Description

A random forest based on conditional inference trees (partykit::ctree()). Calls partykit::cforest() from partykit.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.cforest")

Meta Information

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
replace logical FALSE TRUE, FALSE -
fraction numeric 0.632 [0,1][0, 1]
mtry integer - [0,∞)[0, \infty)
mtryratio numeric - [0,1][0, 1]
applyfun untyped - -
cores integer NULL (−∞,∞)(-\infty, \infty)
trace logical FALSE TRUE, FALSE -
cluster untyped - -
scores untyped - -
teststat character quadratic quadratic, maximum -
splitstat character quadratic quadratic, maximum -
splittest logical FALSE TRUE, FALSE -
testtype character Univariate Bonferroni, MonteCarlo, Univariate, Teststatistic -
nmax untyped - -
pargs untyped - -
alpha numeric 0.05 [0,1][0, 1]
mincriterion numeric 0 [0,1][0, 1]
logmincriterion numeric 0 (−∞,∞)(-\infty, \infty)
minsplit integer 20 [1,∞)[1, \infty)
minbucket integer 7 [1,∞)[1, \infty)
minprob numeric 0.01 [0,1][0, 1]
stump logical FALSE TRUE, FALSE -
lookahead logical FALSE TRUE, FALSE -
MIA logical FALSE TRUE, FALSE -
maxvar integer - [1,∞)[1, \infty)
nresample integer 9999 [1,∞)[1, \infty)
tol numeric 1.490116e-08 [0,∞)[0, \infty)
maxsurrogate integer 0 [0,∞)[0, \infty)
numsurrogate logical FALSE TRUE, FALSE -
maxdepth integer Inf [0,∞)[0, \infty)
multiway logical FALSE TRUE, FALSE -
splittry integer 2 [0,∞)[0, \infty)
intersplit logical FALSE TRUE, FALSE -
majority logical FALSE TRUE, FALSE -
caseweights logical TRUE TRUE, FALSE -
saveinfo logical FALSE TRUE, FALSE -
update logical FALSE TRUE, FALSE -
splitflavour character ctree ctree, exhaustive -
OOB logical FALSE TRUE, FALSE -
simplify logical TRUE TRUE, FALSE -
scale logical TRUE TRUE, FALSE -
nperm integer 1 [0,∞)[0, \infty)
risk character loglik loglik, misclassification -
conditional logical FALSE TRUE, FALSE -
threshold numeric 0.2 (−∞,∞)(-\infty, \infty)

Custom mlr3 parameters

  • mtry:

    • This hyperparameter can alternatively be set via the added hyperparameter mtryratio as mtry = max(ceiling(mtryratio * n_features), 1). Note that mtry and mtryratio are mutually exclusive.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrCForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrCForest$new()

Method oob_error()

The out-of-bag error, calculated using the OOB predictions from partykit.

Usage
LearnerRegrCForest$oob_error()
Returns

numeric(1).

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrCForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sumny

References

Hothorn T, Zeileis A (2015). “partykit: A Modular Toolkit for Recursive Partytioning in R.” Journal of Machine Learning Research, 16(118), 3905-3909. http://jmlr.org/papers/v16/hothorn15a.html.

Hothorn T, Hornik K, Zeileis A (2006). “Unbiased Recursive Partitioning: A Conditional Inference Framework.” Journal of Computational and Graphical Statistics, 15(3), 651–674. doi:10.1198/106186006x133933, https://doi.org/10.1198/106186006x133933.

See Also

Examples

# Define the Learner
learner = lrn("regr.cforest")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Splines Learner

Description

Categorical Regression Splines. Calls crs::crs() from crs.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.crs")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, crs

Parameters

Id Type Default Levels Range
degree integer 3 [0,∞)[0, \infty)
segments integer 1 [1,∞)[1, \infty)
include integer - (−∞,∞)(-\infty, \infty)
lambda untyped - -
lambda.discrete logical FALSE TRUE, FALSE -
lambda.discrete.num integer 100 [0,∞)[0, \infty)
cv character nomad nomad, exhaustive, none -
cv.threshold integer 1000 [0,∞)[0, \infty)
cv.func character cv.ls cv.ls, cv.gcv, cv.aic -
kernel logical TRUE TRUE, FALSE -
degree.max integer 10 [0,∞)[0, \infty)
segments.max integer 10 [1,∞)[1, \infty)
degree.min integer 0 [0,∞)[0, \infty)
segments.min integer 1 [1,∞)[1, \infty)
cv.df.min integer 1 (−∞,∞)(-\infty, \infty)
complexity character degree-knots degree-knots, degree, knots -
knots character quantiles quantiles, uniform, auto -
basis character auto auto, additive, tensor, glp -
prune logical FALSE TRUE, FALSE -
restarts integer 0 [0,∞)[0, \infty)
nmulti integer 5 [0,∞)[0, \infty)
singular.ok logical FALSE TRUE, FALSE -
deriv integer 0 [0,∞)[0, \infty)
data.return logical FALSE TRUE, FALSE -
model.return logical FALSE TRUE, FALSE -
random.seed integer - (−∞,∞)(-\infty, \infty)
tau numeric - [0,1][0, 1]
initial.mesh.size.real untyped - -
initial.mesh.size.integer untyped - -
max.bb.eval untyped - -
min.mesh.size.real untyped - -
min.mesh.size.integer untyped - -
min.frame.size.real untyped - -
min.frame.size.integer untyped - -
display.nomad.progress logical TRUE TRUE, FALSE -
display.warnings logical TRUE TRUE, FALSE -
opts untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrCrs

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrCrs$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrCrs$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

See Also

Examples

# Define the Learner
learner = lrn("regr.crs", cv = "none")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Conditional Inference Tree Learner

Description

Regression Partition Tree where a significance test is used to determine the univariate splits. Calls partykit::ctree() from partykit.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.ctree")

Meta Information

Parameters

Id Type Default Levels Range
teststat character quadratic quadratic, maximum -
splitstat character quadratic quadratic, maximum -
splittest logical FALSE TRUE, FALSE -
testtype character Bonferroni Bonferroni, MonteCarlo, Univariate, Teststatistic -
nmax untyped - -
alpha numeric 0.05 [0,1][0, 1]
mincriterion numeric 0.95 [0,1][0, 1]
logmincriterion numeric - (−∞,∞)(-\infty, \infty)
minsplit integer 20 [1,∞)[1, \infty)
minbucket integer 7 [1,∞)[1, \infty)
minprob numeric 0.01 [0,∞)[0, \infty)
stump logical FALSE TRUE, FALSE -
lookahead logical FALSE TRUE, FALSE -
MIA logical FALSE TRUE, FALSE -
maxvar integer - [1,∞)[1, \infty)
nresample integer 9999 [1,∞)[1, \infty)
tol numeric - [0,∞)[0, \infty)
maxsurrogate integer 0 [0,∞)[0, \infty)
numsurrogate logical FALSE TRUE, FALSE -
mtry integer Inf [0,∞)[0, \infty)
maxdepth integer Inf [0,∞)[0, \infty)
multiway logical FALSE TRUE, FALSE -
splittry integer 2 [0,∞)[0, \infty)
intersplit logical FALSE TRUE, FALSE -
majority logical FALSE TRUE, FALSE -
caseweights logical FALSE TRUE, FALSE -
applyfun untyped - -
cores integer NULL (−∞,∞)(-\infty, \infty)
saveinfo logical TRUE TRUE, FALSE -
update logical FALSE TRUE, FALSE -
splitflavour character ctree ctree, exhaustive -
cluster untyped - -
scores untyped - -
doFit logical TRUE TRUE, FALSE -
maxpts integer 25000 (−∞,∞)(-\infty, \infty)
abseps numeric 0.001 [0,∞)[0, \infty)
releps numeric 0 [0,∞)[0, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrCTree

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrCTree$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrCTree$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sumny

References

Hothorn T, Zeileis A (2015). “partykit: A Modular Toolkit for Recursive Partytioning in R.” Journal of Machine Learning Research, 16(118), 3905-3909. http://jmlr.org/papers/v16/hothorn15a.html.

Hothorn T, Hornik K, Zeileis A (2006). “Unbiased Recursive Partitioning: A Conditional Inference Framework.” Journal of Computational and Graphical Statistics, 15(3), 651–674. doi:10.1198/106186006x133933, https://doi.org/10.1198/106186006x133933.

See Also

Examples

# Define the Learner
learner = lrn("regr.ctree")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Cubist Learner

Description

Rule-based model that is an extension of Quinlan's M5 model tree. Each tree contains linear regression models at the terminal leaves. Calls Cubist::cubist() from Cubist.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.cubist")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, Cubist

Parameters

Id Type Default Levels Range
committees integer - [1,100][1, 100]
unbiased logical FALSE TRUE, FALSE -
rules integer 100 [1,∞)[1, \infty)
extrapolation numeric 100 [0,100][0, 100]
sample integer 0 [0,∞)[0, \infty)
seed integer - (−∞,∞)(-\infty, \infty)
label untyped "outcome" -
neighbors integer - [0,9][0, 9]
strip_time_stamps logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrCubist

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrCubist$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrCubist$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sumny

References

Quinlan, R J, others (1992). “Learning with continuous classes.” In 5th Australian joint conference on artificial intelligence, volume 92, 343–348. World Scientific.

Quinlan, Ross J (1993). “Combining instance-based and model-based learning.” In Proceedings of the tenth international conference on machine learning, 236–243.

See Also

Examples

# Define the Learner
learner = lrn("regr.cubist")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Cross-Validated MCP- or SCAD-penalized Regression Learner

Description

MCP- or SCAD-penalized regression models with k-fold cross validation for choosing the regularization parameter lambda. Calls ncvreg::cv.ncvreg() from package ncvreg with family set to "gaussian" (default). Poisson family is also supported. Covariates are internally standardized by ncvreg prior to model fitting.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.cv_ncvreg")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, ncvreg

Parameters

Id Type Default Levels Range
family character gaussian gaussian, poisson -
penalty character MCP MCP, SCAD, lasso -
gamma numeric - [0,∞)[0, \infty)
alpha numeric 1 [0,1][0, 1]
lambda.min numeric - [0,∞)[0, \infty)
nlambda integer 100 [1,∞)[1, \infty)
lambda untyped - -
eps numeric 1e-04 (−∞,∞)(-\infty, \infty)
max.iter integer 10000 [1,∞)[1, \infty)
convex logical TRUE TRUE, FALSE -
dfmax integer - [0,∞)[0, \infty)
penalty.factor untyped - -
warn logical TRUE TRUE, FALSE -
returnX logical TRUE TRUE, FALSE -
cluster untyped - -
nfolds integer 10 (−∞,∞)(-\infty, \infty)
fold untyped - -
returnY logical FALSE TRUE, FALSE -
trace logical FALSE TRUE, FALSE -
pred_lambda numeric - (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrNCVreg

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrNCVreg$new()

Method selected_features()

Returns the set of selected features which have non-zero coefficients. Calls the internal coef.cv.ncvreg() function.

Usage
LearnerRegrNCVreg$selected_features(lambda = NULL)
Arguments
lambda

(numeric(1))
Custom lambda, defaults to the lambda with the minimum CV error.

Returns

(character()) vector of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrNCVreg$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Breheny P, Huang J (2011). “Coordinate descent algorithms for nonconvex penalized regression, with applications to biological feature selection.” Annals of Applied Statistics, 5(1), 232–253. ISSN 1932-6157, doi:10.1214/10-AOAS388.

See Also

Examples

# Define the Learner
learner = lrn("regr.cv_ncvreg")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Decision Stump Learner

Description

Decision Stump Learner. Calls RWeka::DecisionStump() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.decision_stump")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrDecisionStump

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrDecisionStump$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrDecisionStump$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrDecisionStump$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrDecisionStump$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("regr.decision_stump")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Decision Table Learner

Description

Simple Decision Table majority regressor. Calls RWeka::make_Weka_classifier() from RWeka.

Initial parameter values

  • E:

    • Has only 2 out of 4 original evaluation measures : rmse and mae with rmse being the default

    • Reason for change: this learner should only contain evaluation measures appropriate for regression tasks

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • P_best:

    • original id: P

  • D_best:

    • original id: D

  • N_best:

    • original id: N

  • S_best:

    • original id: S

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.decision_table")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
S character BestFirst BestFirst, GreedyStepwise -
X integer 1 (−∞,∞)(-\infty, \infty)
E character rmse rmse, mae -
I logical - TRUE, FALSE -
R logical - TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
P_best untyped - -
D_best character 1 0, 1, 2 -
N_best integer - (−∞,∞)(-\infty, \infty)
S_best integer 1 (−∞,∞)(-\infty, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrDecisionTable

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrDecisionTable$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrDecisionTable$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrDecisionTable$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrDecisionTable$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Kohavi R (1995). “The Power of Decision Tables.” In 8th European Conference on Machine Learning, 174–189.

See Also

Examples

# Define the Learner
learner = lrn("regr.decision_table")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Implementation of Multivariate Adaptive Regression Splines

Description

This is an alternative implementation of MARS (Multivariate Adaptive Regression Splines). MARS is trademarked and thus not used as the name. The name "earth" stands for "Enhanced Adaptive Regression Through Hinges".

Calls earth::earth() from earth.

Details

Methods for variance estimations are not yet implemented.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.earth")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, earth

Parameters

Id Type Default Levels Range
wp untyped NULL -
offset untyped NULL -
keepxy logical FALSE TRUE, FALSE -
trace character 0 0, .3, .5, 1, 2, 3, 4, 5 -
degree integer 1 [1,∞)[1, \infty)
penalty numeric 2 [−1,∞)[-1, \infty)
nk untyped NULL -
thresh numeric 0.001 (−∞,∞)(-\infty, \infty)
minspan numeric 0 [0,∞)[0, \infty)
endspan numeric 0 [0,∞)[0, \infty)
newvar.penalty numeric 0 [0,∞)[0, \infty)
fast.k integer 20 [0,∞)[0, \infty)
fast.beta integer 1 [0,1][0, 1]
linpreds untyped FALSE -
allowed untyped - -
pmethod character backward backward, none, exhaustive, forward, seqrep, cv -
nprune integer - [0,∞)[0, \infty)
nfold integer 0 [0,∞)[0, \infty)
ncross integer 1 [0,∞)[0, \infty)
stratify logical TRUE TRUE, FALSE -
varmod.method character none none, const, lm, rlm, earth, gam, power, power0, x.lm, x.rlm, [...] -
varmod.exponent numeric 1 (−∞,∞)(-\infty, \infty)
varmod.conv numeric 1 [0,1][0, 1]
varmod.clamp numeric 0.1 (−∞,∞)(-\infty, \infty)
varmod.minspan numeric -3 (−∞,∞)(-\infty, \infty)
Scale.y logical FALSE TRUE, FALSE -
Adjust.endspan numeric 2 (−∞,∞)(-\infty, \infty)
Auto.linpreds logical TRUE TRUE, FALSE -
Force.weights logical FALSE TRUE, FALSE -
Use.beta.cache logical TRUE TRUE, FALSE -
Force.xtx.prune logical FALSE TRUE, FALSE -
Get.leverages logical TRUE TRUE, FALSE -
Exhaustive.tol numeric 1e-10 (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrEarth

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrEarth$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrEarth$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

pkopper

References

Milborrow, Stephen, Hastie, T, Tibshirani, R (2014). “Earth: multivariate adaptive regression spline models.” R package version, 3(7).

Friedman, H J (1991). “Multivariate adaptive regression splines.” The annals of statistics, 19(1), 1–67.

See Also

Examples

# Define the Learner
learner = lrn("regr.earth")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Decision Tree Learner

Description

Evolutionary learning of globally optimal regression trees. Calls evtree::evtree() fromevtree.

Initial parameter values

pmutatemajor, pmutateminor, pcrossover, psplit, and pprune, are scaled internally to sum to 100.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrEvtree

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrEvtree$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrEvtree$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

See Also

Linear Regression with Exhaustive Search

Description

Linear Regression, where the features used in the model are selected by running Exhaustive Search.

Calls ExhaustiveSearch::ExhaustiveSearch() from ExhaustiveSearch.

Initial parameter values

  • family:

    • Actual default: NULL

    • Adjusted default: "gaussian"

    • Reason for change: To comply with mlr3 architecture, we differentiate between classification and regression learners.

  • nThreads:

    • Actual default: NULL

    • Adjusted default: 1

    • Reason for change: Suppressing the automatic internal parallelization if cv.folds > 0.

  • quietly:

    • Actual default: FALSE

    • Adjusted default: TRUE

    • Reason for change: Suppression of constant printing to console

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.exhaustive_search")

Meta Information

Parameters

Id Type Default Levels Range
family character - gaussian, binomial -
performanceMeasure character - MSE, AIC -
combsUpTo integer - [1,∞)[1, \infty)
nResults integer 5000 [1,∞)[1, \infty)
nThreads integer - [1,∞)[1, \infty)
testSetIDs integer - [1,∞)[1, \infty)
errorVal untyped -1 -
quietly logical - TRUE, FALSE -
checkLarge logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrExhaustiveSearch

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrExhaustiveSearch$new()

Method selected_features()

Extracts selected features of this learner.

Usage
LearnerRegrExhaustiveSearch$selected_features()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrExhaustiveSearch$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

ngerman

See Also

Examples

# define learner
learner = lrn("regr.exhaustive_search", predict_type = "se")

# define task
tsk_cars = tsk("mtcars")

# train learner
learner$train(tsk_cars)

# extract selected features
learner$selected_features()

# predict on training task
learner$predict(tsk_cars)

Regression Fast Nearest Neighbor Search Learner

Description

Fast Nearest Neighbour Regression. Calls FNN::knn.reg() from FNN.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.fnn")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, FNN

Parameters

Id Type Default Levels Range
k integer 1 [1,∞)[1, \infty)
algorithm character kd_tree kd_tree, cover_tree, brute -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrFNN

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrFNN$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrFNN$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Boltz, Sylvain, Debreuve, Eric, Barlaud, Michel (2007). “kNN-based high-dimensional Kullback-Leibler distance for tracking.” In Eighth International Workshop on Image Analysis for Multimedia Interactive Services (WIAMIS'07), 16–16. IEEE.

See Also

Examples

# Define the Learner
learner = lrn("regr.fnn")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Fuzzy Rule-based System Learner

Description

Fuzzy rule-based regression learner supporting multiple inference engines such as WM, HYFIS, ANFIS, DENFIS, SBC, and several genetic fuzzy systems. Calls frbs::frbs.learn() from frbs.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.frbs")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, frbs

Parameters

Id Type Default Levels Range
alpha.heuristic numeric 1 [2.22044604925031e−16,∞)[2.22044604925031e-16, \infty)
d numeric 2 (−∞,∞)(-\infty, \infty)
Dthr numeric 0.1 [0,1][0, 1]
eps.high numeric 0.5 [2.22044604925031e−16,∞)[2.22044604925031e-16, \infty)
eps.low numeric 0.15 [2.22044604925031e−16,∞)[2.22044604925031e-16, \infty)
epsilon numeric 0.9 [0,1][0, 1]
max.gen integer 10 [1,∞)[1, \infty)
max.iter integer 10 [1,∞)[1, \infty)
max.tune integer 10 [1,∞)[1, \infty)
method.type character WM WM, SBC, HYFIS, ANFIS, DENFIS, FS.HGD, FIR.DM, GFS.FR.MOGUL, GFS.THRIFT, GFS.LT.RS -
mode.tuning character GLOBAL GLOBAL, LOCAL -
num.labels integer 7 [1,∞)[1, \infty)
persen_cross numeric 1 [0,1][0, 1]
persen_mutant numeric 1 [0,1][0, 1]
popu.size integer 10 [1,∞)[1, \infty)
r.a numeric 0.5 [2.22044604925031e−16,∞)[2.22044604925031e-16, \infty)
range.data untyped NULL -
rule.selection logical FALSE TRUE, FALSE -
step.size numeric 0.01 [0,1][0, 1]
type.defuz character WAM WAM, FIRST.MAX, LAST.MAX, MEAN.MAX, COG -
type.implication.func character ZADEH DIENES_RESHER, LUKASIEWICZ, ZADEH, GOGUEN, GODEL, SHARP, MIZUMOTO, DUBOIS_PRADE, MIN -
type.mf character GAUSSIAN TRIANGLE, TRAPEZOID, GAUSSIAN, SIGMOID, BELL -
type.snorm character MAX MAX, HAMACHER, YAGER, PRODUCT, BOUNDED -
type.tnorm character MIN MIN, HAMACHER, YAGER, PRODUCT, BOUNDED -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrFrbs

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrFrbs$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrFrbs$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Riza LS, Bergmeir C, Herrera F, Benitez JM (2015). “frbs: Fuzzy Rule-Based Systems for Classification and Regression in R.” Journal of Statistical Software, 65(6), 1–30. http://www.jstatsoft.org/v65/i06/.

See Also

Examples

# Define the Learner
learner = lrn("regr.frbs")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Generalized Additive Model Learner

Description

Generalized additive models. Calls mgcv::gam() from package mgcv.

Formula

A gam formula specific to the task at hand is required for the formula parameter (see example and ?mgcv::formula.gam). Beware, if no formula is provided, a fallback formula is used that will make the gam behave like a glm (this behavior is required for the unit tests). Only features specified in the formula will be used, superseding columns with col_roles "feature" in the task.

Offset

If a Task contains a column with the offset role, it is automatically incorporated during training via the offset argument in mgcv::gam(). No offset is applied during prediction for this learner.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.gam")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, mgcv

Parameters

Id Type Default Levels Range
family character gaussian gaussian, poisson -
formula untyped - -
method character GCV.Cp GCV.Cp, GACV.Cp, REML, P-REML, ML, P-ML -
optimizer untyped c("outer", "newton") -
scale numeric 0 (−∞,∞)(-\infty, \infty)
select logical FALSE TRUE, FALSE -
knots untyped NULL -
sp untyped NULL -
min.sp untyped NULL -
H untyped NULL -
gamma numeric 1 [1,∞)[1, \infty)
paraPen untyped NULL -
G untyped NULL -
in.out untyped NULL -
drop.unused.levels logical TRUE TRUE, FALSE -
drop.intercept logical FALSE TRUE, FALSE -
nei untyped - -
nthreads integer 1 [1,∞)[1, \infty)
irls.reg numeric 0 [0,∞)[0, \infty)
epsilon numeric 1e-07 [0,∞)[0, \infty)
maxit integer 200 (−∞,∞)(-\infty, \infty)
trace logical FALSE TRUE, FALSE -
mgcv.tol numeric 1e-07 [0,∞)[0, \infty)
mgcv.half integer 15 [0,∞)[0, \infty)
rank.tol numeric 1.490116e-08 [0,∞)[0, \infty)
nlm untyped list() -
optim untyped list() -
newton untyped list() -
idLinksBases logical TRUE TRUE, FALSE -
scalePenalty logical TRUE TRUE, FALSE -
efs.lspmax integer 15 [0,∞)[0, \infty)
efs.tol numeric 0.1 [0,∞)[0, \infty)
scale.est character fletcher fletcher, pearson, deviance -
ncv.threads integer 1 [1,∞)[1, \infty)
edge.correct logical FALSE TRUE, FALSE -
block.size integer 1000 (−∞,∞)(-\infty, \infty)
unconditional logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrGam

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrGam$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrGam$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

pierrecamilleri

References

Hastie, J T, Tibshirani, J R (2017). Generalized additive models. Routledge.

Wood, Simon (2012). “mgcv: Mixed GAM Computation Vehicle with GCV/AIC/REML smoothness estimation.”

Examples

# simple example
t = tsk("mtcars")
l = lrn("regr.gam")
l$param_set$values$formula = mpg ~ cyl + am + s(disp) + s(hp)
l$train(t)
l$model

Boosted Generalized Additive Regression Learner

Description

Fit a generalized additive regression model using a boosting algorithm. Calls mboost::gamboost() from mboost.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.gamboost")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, mboost

Parameters

Id Type Default Levels Range
baselearner character bbs bbs, bols, btree -
dfbase integer 4 (−∞,∞)(-\infty, \infty)
family character Gaussian Gaussian, Laplace, Huber, Poisson, GammaReg, NBinomial, Hurdle, custom -
custom.family untyped - -
nuirange untyped c(0, 100) -
d numeric NULL (−∞,∞)(-\infty, \infty)
mstop integer 100 (−∞,∞)(-\infty, \infty)
nu numeric 0.1 (−∞,∞)(-\infty, \infty)
risk character inbag inbag, oobag, none -
oobweights untyped NULL -
trace logical FALSE TRUE, FALSE -
stopintern untyped FALSE -
na.action untyped stats::na.omit -

Offset

If a Task contains a column with the offset role, it is automatically incorporated via the offset argument in mboost's training function. No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrGAMBoost

Methods

Public methods

Inherited methods

Method new()

Create a LearnerRegrGAMBoost object.

Usage
LearnerRegrGAMBoost$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrGAMBoost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Bühlmann, Peter, Yu, Bin (2003). “Boosting with the L 2 loss: regression and classification.” Journal of the American Statistical Association, 98(462), 324–339.

See Also

Examples

lrn("regr.gamboost")

Regression Gaussian Processes Learner From Weka

Description

Gaussian Processes. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • E_poly:

    • original id: E

  • L_poly:

    • original id: L (duplicated L for when K is set to PolyKernel)

  • C_poly:

    • original id: C

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

  • output-debug-info for kernel parameter removed:

    • enables debugging output (if available) to be printed

  • Reason for change: The parameter is removed because it's unclear how to actually use it.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.gaussian_processes")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
L numeric 1 (−∞,∞)(-\infty, \infty)
N character 0 0, 1, 2 -
K character supportVector.PolyKernel supportVector.NormalizedPolyKernel, supportVector.PolyKernel, supportVector.Puk, supportVector.RBFKernel, supportVector.StringKernel -
S integer 1 (−∞,∞)(-\infty, \infty)
E_poly numeric 1 (−∞,∞)(-\infty, \infty)
L_poly logical FALSE TRUE, FALSE -
C_poly integer 250007 (−∞,∞)(-\infty, \infty)
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrGaussianProcesses

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrGaussianProcesses$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrGaussianProcesses$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrGaussianProcesses$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrGaussianProcesses$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Mackay DJ (1998). “Introduction to Gaussian Processes.”

See Also

Examples

# Define the Learner
learner = lrn("regr.gaussian_processes")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Gaussian Process Learner

Description

Gaussian process for regression. Calls kernlab::gausspr() from kernlab. Parameters sigma, degree, scale, offset and order are added to make tuning kpar easier. If kpar is provided then these new parameters are ignored. If none are provided then the default "automatic" is used for kpar.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.gausspr")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, kernlab

Parameters

Id Type Default Levels Range
scaled untyped TRUE -
kernel character rbfdot rbfdot, polydot, vanilladot, tanhdot, laplacedot, besseldot, anovadot, splinedot -
sigma numeric - (−∞,∞)(-\infty, \infty)
degree numeric - (−∞,∞)(-\infty, \infty)
scale numeric - (−∞,∞)(-\infty, \infty)
offset numeric - (−∞,∞)(-\infty, \infty)
order numeric - (−∞,∞)(-\infty, \infty)
kpar untyped "automatic" -
var numeric 0.001 [0.001,∞)[0.001, \infty)
variance.model logical FALSE TRUE, FALSE -
tol numeric 0.001 [0,∞)[0, \infty)
fit logical TRUE TRUE, FALSE -
na.action untyped na.omit -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrGausspr

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrGausspr$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrGausspr$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Karatzoglou, Alexandros, Smola, Alex, Hornik, Kurt, Zeileis, Achim (2004). “kernlab-an S4 package for kernel methods in R.” Journal of statistical software, 11(9), 1–20.

See Also

Examples

# Define the Learner
learner = lrn("regr.gausspr")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Gradient Boosting Machine Learner

Description

Gradient Boosting Regression Algorithm. Calls gbm::gbm() from gbm.

Weights are ignored for quantile prediction.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.gbm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “quantiles”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, gbm

Parameters

Id Type Default Levels Range
distribution character gaussian gaussian, laplace, poisson, tdist -
n.trees integer 100 [1,∞)[1, \infty)
interaction.depth integer 1 [1,∞)[1, \infty)
n.minobsinnode integer 10 [1,∞)[1, \infty)
shrinkage numeric 0.001 [0,∞)[0, \infty)
bag.fraction numeric 0.5 [0,1][0, 1]
train.fraction numeric 1 [0,1][0, 1]
cv.folds integer 0 (−∞,∞)(-\infty, \infty)
keep.data logical FALSE TRUE, FALSE -
verbose logical FALSE TRUE, FALSE -
n.cores integer 1 (−∞,∞)(-\infty, \infty)
var.monotone untyped - -

Parameter changes

  • keep.data:

    • Actual default: TRUE

    • Adjusted default: FALSE

    • Reason for change: keep.data = FALSE saves memory during model fitting.

  • n.cores:

    • Actual default: NULL

    • Adjusted default: 1

    • Reason for change: Suppressing the automatic internal parallelization if cv.folds > 0.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrGBM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrGBM$new()

Method importance()

The importance scores are extracted by gbm::relative.influence() from the model.

Usage
LearnerRegrGBM$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrGBM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Friedman, H J (2002). “Stochastic gradient boosting.” Computational statistics & data analysis, 38(4), 367–378.

See Also

Examples

lrn("regr.gbm")

Generalized Linear Regression

Description

Generalized linear model. Calls stats::glm() from base package 'stats'. For logistic regression please use mlr_learners_classif.log_reg.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.glm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, 'stats'

Parameters

Id Type Default Levels Range
singular.ok logical TRUE TRUE, FALSE -
x logical FALSE TRUE, FALSE -
y logical TRUE TRUE, FALSE -
model logical TRUE TRUE, FALSE -
etastart untyped - -
mustart untyped - -
start untyped NULL -
family character gaussian gaussian, poisson, quasipoisson, Gamma, inverse.gaussian -
na.action character - na.omit, na.pass, na.fail, na.exclude -
link character - logit, probit, cauchit, cloglog, identity, log, sqrt, 1/mu^2, inverse -
epsilon numeric 1e-08 (−∞,∞)(-\infty, \infty)
maxit numeric 25 (−∞,∞)(-\infty, \infty)
trace logical FALSE TRUE, FALSE -
dispersion untyped NULL -
type character link response, link, terms -
use_pred_offset logical TRUE TRUE, FALSE -

Initial parameter values

  • type

    • Actual default: "link"

    • Adjusted default: "response"

    • Reason for change: Response scale more natural for predictions.

Offset

If a Task has a column with the role offset, it will automatically be used during training. The offset is incorporated through the formula interface to ensure compatibility with stats::glm(). We add it to the model formula as ⁠offset(<column_name>)⁠ and also include it in the training data. During prediction, the default behavior is to use the offset column from the test set (enabled by use_pred_offset = TRUE). Otherwise, if the user sets use_pred_offset = FALSE, a zero offset is applied, effectively disabling the offset adjustment during prediction.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrGlm

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrGlm$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrGlm$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

salauer

References

Hosmer Jr, W D, Lemeshow, Stanley, Sturdivant, X R (2013). Applied logistic regression, volume 398. John Wiley & Sons.

See Also

Examples

# Define the Learner
learner = lrn("regr.glm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Boosted Generalized Linear Regression Learner

Description

Fit a generalized linear regression model using a boosting algorithm. Calls mboost::glmboost() from mboost.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.glmboost")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, mboost

Parameters

Id Type Default Levels Range
family character Gaussian Gaussian, Laplace, Huber, Poisson, GammaReg, NBinomial, Hurdle, custom -
custom.family untyped - -
nuirange untyped c(0, 100) -
d numeric NULL (−∞,∞)(-\infty, \infty)
center logical TRUE TRUE, FALSE -
mstop integer 100 (−∞,∞)(-\infty, \infty)
nu numeric 0.1 (−∞,∞)(-\infty, \infty)
risk character inbag inbag, oobag, none -
oobweights untyped NULL -
trace logical FALSE TRUE, FALSE -
stopintern untyped FALSE -
na.action untyped stats::na.omit -
contrasts.arg untyped - -

Offset

If a Task contains a column with the offset role, it is automatically incorporated via the offset argument in mboost's training function. No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrGLMBoost

Methods

Public methods

Inherited methods

Method new()

Create a LearnerRegrGLMBoost object.

Usage
LearnerRegrGLMBoost$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrGLMBoost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Bühlmann, Peter, Yu, Bin (2003). “Boosting with the L 2 loss: regression and classification.” Journal of the American Statistical Association, 98(462), 324–339.

See Also

Examples

# Define the Learner
learner = lrn("regr.glmboost")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Gaussian Process Learner (GPfit)

Description

Gaussian process regression via GPfit::GP_fit() from GPfit.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.gpfit")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, GPfit

Parameters

Id Type Default Levels Range
control untyped NULL -
nug_thres numeric - [0,∞)[0, \infty)
trace logical - TRUE, FALSE -
maxit integer - [1,∞)[1, \infty)
optim_start untyped - -
scale logical - TRUE, FALSE -
type character exponential exponential, matern -
matern_nu_k integer 0 [0,∞)[0, \infty)
power numeric 1.95 [1,2][1, 2]

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrGPfit

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrGPfit$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrGPfit$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

As the optimization routine assumes that the inputs are scaled to the unit hypercube, the input gets scaled for each variable by default. If this is not wanted, ⁠scale = FALSE`` has to be set. We replace the GPfit parameter ⁠corr = list(type = 'exponential',power = 1.95)⁠to be separate parameters⁠typeandpower⁠, in the case of ⁠corr = list(type = 'matern', nu = 0.5)⁠, the separate parameters are ⁠typeandmatern_nu_k = 0⁠, and nu is computed by ⁠nu = (2 * matern_nu_k + 1) / 2 = 0.5'.

Author(s)

awinterstetter

References

MacDonald, B., Ranjan, P., Chipman, H. (2015). “GPfit: An R package for fitting a Gaussian process model to deterministic simulator outputs.” Journal of Statistical Software, 64, 1–23. ISSN 1548-7660, doi:10.18637/jss.v064.i12.

See Also

Examples

# Define the Learner
learner = lrn("regr.gpfit", maxit = 1)
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression H2O Deep Learning Learner

Description

Regression feed-forward multilayer artificial neural network learner. Class h2o::h2o.deeplearning() from package h2o.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.h2o.deeplearning")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, h2o

Parameters

Id Type Default Levels Range
activation character Rectifier Rectifier, Tanh, TanhWithDropout, RectifierWithDropout, Maxout, MaxoutWithDropout -
adaptive_rate logical TRUE TRUE, FALSE -
autoencoder logical FALSE TRUE, FALSE -
average_activation numeric 0 (−∞,∞)(-\infty, \infty)
categorical_encoding character AUTO AUTO, Enum, OneHotInternal, OneHotExplicit, Binary, Eigen, LabelEncoder, SortByResponse, EnumLimited -
checkpoint untyped NULL -
diagnostics logical TRUE TRUE, FALSE -
distribution character AUTO AUTO, gaussian, poisson, gamma, tweedie, laplace, huber, quantile -
elastic_averaging logical FALSE TRUE, FALSE -
elastic_averaging_moving_rate numeric 0.9 (−∞,∞)(-\infty, \infty)
elastic_averaging_regularization numeric 0.001 (−∞,∞)(-\infty, \infty)
epochs numeric 10 [1,∞)[1, \infty)
epsilon numeric 1e-08 [1e−10,1e−04][1e-10, 1e-04]
export_checkpoints_dir untyped NULL -
export_weights_and_biases logical FALSE TRUE, FALSE -
fast_mode logical TRUE TRUE, FALSE -
force_load_balance logical TRUE TRUE, FALSE -
hidden untyped c(200L, 200L) -
hidden_dropout_ratios numeric 0.5 (−∞,∞)(-\infty, \infty)
huber_alpha numeric 0.9 [0,1][0, 1]
ignore_const_cols logical TRUE TRUE, FALSE -
initial_weight_distribution character UniformAdaptive UniformAdaptive, Uniform, Normal -
initial_weight_scale numeric 1 (−∞,∞)(-\infty, \infty)
input_dropout_ratio numeric 0 (−∞,∞)(-\infty, \infty)
l1 numeric 0 (−∞,∞)(-\infty, \infty)
l2 numeric 0 (−∞,∞)(-\infty, \infty)
loss character Automatic Automatic, Quantile, Quadratic, Absolute, Huber -
max_categorical_features integer 2147483647 [1,∞)[1, \infty)
max_runtime_secs numeric 0 [0,∞)[0, \infty)
max_w2 numeric 3.402823e+38 (−∞,∞)(-\infty, \infty)
mini_batch_size integer 1 (−∞,∞)(-\infty, \infty)
missing_values_handling character MeanImputation MeanImputation, Skip -
momentum_ramp numeric 1e+06 (−∞,∞)(-\infty, \infty)
momentum_stable numeric 0 (−∞,∞)(-\infty, \infty)
momentum_start numeric 0 (−∞,∞)(-\infty, \infty)
nesterov_accelerated_gradient logical TRUE TRUE, FALSE -
overwrite_with_best_model logical TRUE TRUE, FALSE -
pretrained_autoencoder untyped NULL -
quantile_alpha numeric 0.5 [0,1][0, 1]
quiet_mode logical FALSE TRUE, FALSE -
rate numeric 0.005 [0,1][0, 1]
rate_annealing numeric 1e-06 [0,∞)[0, \infty)
rate_decay numeric 1 [0,∞)[0, \infty)
regression_stop numeric 1e-06 [−1,∞)[-1, \infty)
replicate_training_data logical TRUE TRUE, FALSE -
reproducible logical FALSE TRUE, FALSE -
rho numeric 0.99 [0,∞)[0, \infty)
score_duty_cycle numeric 0.1 (−∞,∞)(-\infty, \infty)
score_each_iteration logical FALSE TRUE, FALSE -
score_interval numeric 5 (−∞,∞)(-\infty, \infty)
score_training_samples integer 10000 (−∞,∞)(-\infty, \infty)
score_validation_samples integer 0 (−∞,∞)(-\infty, \infty)
seed integer -1 (−∞,∞)(-\infty, \infty)
shuffle_training_data logical FALSE TRUE, FALSE -
single_node_mode logical FALSE TRUE, FALSE -
sparse logical FALSE TRUE, FALSE -
sparsity_beta numeric 0 (−∞,∞)(-\infty, \infty)
standardize logical TRUE TRUE, FALSE -
stopping_metric character AUTO AUTO, deviance, MSE, RMSE, MAE, RMSLE -
stopping_rounds integer 5 [0,∞)[0, \infty)
stopping_tolerance numeric 0 [0,∞)[0, \infty)
target_ratio_comm_to_comp numeric 0.05 (−∞,∞)(-\infty, \infty)
train_samples_per_iteration integer -2 [−2,∞)[-2, \infty)
tweedie_power numeric 1.5 [1,2][1, 2]
verbose logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrH2ODeeplearning

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrH2ODeeplearning$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrH2ODeeplearning$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Fryda T, LeDell E, Gill N, Aiello S, Fu A, Candel A, Click C, Kraljevic T, Nykodym T, Aboyoun P, Kurka M, Malohlava M, Poirier S, Wong W (2025). h2o: R Interface for the 'H2O' Scalable Machine Learning Platform. R package version 3.46.0.9, https://github.com/h2oai/h2o-3.

See Also

Examples

# Define the Learner
learner = lrn("regr.h2o.deeplearning")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression H2O GBM Learner

Description

Gradient boosting machine learner. Class h2o::h2o.gbm() from package h2o.

H2O Connection

If no running H2O connection is found, the learner will automatically start a local H2O server on ⁠127.0.0.1⁠ via h2o::h2o.init(). If you want to connect to a remote H2O cluster, call h2o::h2o.init() with the appropriate arguments before training or predicting.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.h2o.gbm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, h2o

Parameters

Id Type Default Levels Range
auto_rebalance logical TRUE TRUE, FALSE -
build_tree_one_node logical FALSE TRUE, FALSE -
categorical_encoding character AUTO AUTO, Enum, OneHotInternal, OneHotExplicit, Binary, Eigen, LabelEncoder, SortByResponse, EnumLimited -
check_constant_response logical TRUE TRUE, FALSE -
checkpoint untyped NULL -
col_sample_rate numeric 1 [0,1][0, 1]
col_sample_rate_change_per_level numeric 1 [0,2][0, 2]
col_sample_rate_per_tree numeric 1 [0,1][0, 1]
custom_distribution_func untyped NULL -
distribution character gaussian AUTO, poisson, laplace, tweedie, gaussian, huber, gamma, quantile, custom -
export_checkpoints_dir untyped NULL -
histogram_type character AUTO AUTO, UniformAdaptive, Random, QuantilesGlobal, RoundRobin, UniformRobust -
huber_alpha numeric 0.9 [0,1][0, 1]
ignore_const_cols logical TRUE TRUE, FALSE -
in_training_checkpoints_dir untyped NULL -
in_training_checkpoints_tree_interval integer 1 [1,∞)[1, \infty)
interaction_constraints untyped NULL -
learn_rate numeric 0.1 [0,1][0, 1]
learn_rate_annealing numeric 1 [0,1][0, 1]
max_abs_leafnode_pred numeric Inf [0,∞)[0, \infty)
max_depth integer 5 [1,∞)[1, \infty)
max_runtime_secs numeric 0 [0,∞)[0, \infty)
min_rows integer 10 [1,∞)[1, \infty)
min_split_improvement numeric 1e-05 [0,∞)[0, \infty)
monotone_constraints untyped NULL -
nbins integer 20 [1,∞)[1, \infty)
nbins_cats integer 1024 [1,∞)[1, \infty)
nbins_top_level integer 1024 [1,∞)[1, \infty)
ntrees integer 50 [1,∞)[1, \infty)
pred_noise_bandwidth numeric 0 [0,∞)[0, \infty)
quantile_alpha numeric 0.5 [0,1][0, 1]
sample_rate numeric 1 [0,1][0, 1]
score_each_iteration logical FALSE TRUE, FALSE -
score_tree_interval integer 0 [0,∞)[0, \infty)
seed integer -1 (−∞,∞)(-\infty, \infty)
stopping_metric character AUTO AUTO, deviance, MSE, RMSE, MAE, RMSLE -
stopping_rounds integer 0 [0,∞)[0, \infty)
stopping_tolerance numeric 0.001 [0,∞)[0, \infty)
tweedie_power numeric 1.5 [1,2][1, 2]
verbose logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrH2OGBM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrH2OGBM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrH2OGBM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Fryda T, LeDell E, Gill N, Aiello S, Fu A, Candel A, Click C, Kraljevic T, Nykodym T, Aboyoun P, Kurka M, Malohlava M, Poirier S, Wong W (2025). h2o: R Interface for the 'H2O' Scalable Machine Learning Platform. R package version 3.46.0.9, https://github.com/h2oai/h2o-3.

See Also

Examples

# Define the Learner
learner = lrn("regr.h2o.gbm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Generalized Linear Regression H2O Learner

Description

Generalized linear model. Calls h2o::h2o.glm() from package h2o.

H2O Connection

If no running H2O connection is found, the learner will automatically start a local H2O server on ⁠127.0.0.1⁠ via h2o::h2o.init(). If you want to connect to a remote H2O cluster, call h2o::h2o.init() with the appropriate arguments before training or predicting.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.h2o.glm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, h2o

Parameters

Id Type Default Levels Range
alpha numeric 0.5 [0,1][0, 1]
beta_constraints untyped NULL -
beta_epsilon numeric 1e-04 [0,∞)[0, \infty)
build_null_model logical FALSE TRUE, FALSE -
calc_like logical FALSE TRUE, FALSE -
checkpoint untyped NULL -
cold_start logical FALSE TRUE, FALSE -
compute_p_values logical FALSE TRUE, FALSE -
early_stopping logical TRUE TRUE, FALSE -
export_checkpoints_dir untyped NULL -
generate_scoring_history logical FALSE TRUE, FALSE -
generate_variable_inflation_factors logical FALSE TRUE, FALSE -
gradient_epsilon numeric -1 [0,∞)[0, \infty)
HGLM logical FALSE TRUE, FALSE -
ignore_const_cols logical TRUE TRUE, FALSE -
interactions untyped NULL -
interaction_pairs untyped NULL -
intercept logical TRUE TRUE, FALSE -
lambda numeric 1e-05 [0,∞)[0, \infty)
lambda_min_ratio numeric -1 [0,1][0, 1]
lambda_search logical FALSE TRUE, FALSE -
link character family_default family_default, identity, log, inverse -
max_active_predictors integer -1 [1,∞)[1, \infty)
max_iterations integer -1 [0,∞)[0, \infty)
max_runtime_secs numeric 0 [0,∞)[0, \infty)
missing_values_handling character MeanImputation MeanImputation, Skip, PlugValues -
nlambdas integer -1 [1,∞)[1, \infty)
non_negative logical FALSE TRUE, FALSE -
objective_epsilon numeric -1 [0,∞)[0, \infty)
obj_reg numeric -1 [0,∞)[0, \infty)
plug_values untyped NULL -
random_columns untyped NULL -
rand_family untyped NULL -
rand_link untyped NULL -
remove_collinear_columns logical FALSE TRUE, FALSE -
score_each_iteration logical FALSE TRUE, FALSE -
score_iteration_interval integer -1 (−∞,∞)(-\infty, \infty)
seed integer -1 (−∞,∞)(-\infty, \infty)
solver character AUTO AUTO, IRLSM, L_BFGS, COORDINATE_DESCENT_NAIVE, COORDINATE_DESCENT, GRADIENT_DESCENT_LH, GRADIENT_DESCENT_SQERR -
standardize logical TRUE TRUE, FALSE -
startval untyped NULL -
stopping_metric character AUTO AUTO, deviance, MSE, RMSE, MAE, RMSLE -
stopping_rounds integer 0 [0,∞)[0, \infty)
stopping_tolerance numeric 0.001 [0,∞)[0, \infty)

Model family

This learner calls h2o::h2o.glm() with family = "gaussian"

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrH2OGLM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrH2OGLM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrH2OGLM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Fryda T, LeDell E, Gill N, Aiello S, Fu A, Candel A, Click C, Kraljevic T, Nykodym T, Aboyoun P, Kurka M, Malohlava M, Poirier S, Wong W (2025). h2o: R Interface for the 'H2O' Scalable Machine Learning Platform. R package version 3.46.0.9, https://github.com/h2oai/h2o-3.

See Also

Examples

# Define the Learner
learner = lrn("regr.h2o.glm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression H2O Random Forest Learner

Description

Regression random forest learner. Calls h2o::h2o.randomForest() from package h2o.

H2O Connection

If no running H2O connection is found, the learner will automatically start a local H2O server on ⁠127.0.0.1⁠ via h2o::h2o.init(). If you want to connect to a remote H2O cluster, call h2o::h2o.init() with the appropriate arguments before training or predicting.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.h2o.randomForest")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, h2o

Parameters

Id Type Default Levels Range
build_tree_one_node logical FALSE TRUE, FALSE -
categorical_encoding character AUTO AUTO, Enum, OneHotInternal, OneHotExplicit, Binary, Eigen, LabelEncoder, SortByResponse, EnumLimited -
check_constant_response logical TRUE TRUE, FALSE -
checkpoint untyped NULL -
col_sample_rate_change_per_level numeric 1 [0,2][0, 2]
col_sample_rate_per_tree numeric 1 [0,1][0, 1]
export_checkpoints_dir untyped NULL -
histogram_type character AUTO AUTO, UniformAdaptive, Random, QuantilesGlobal, RoundRobin, UniformRobust -
ignore_const_cols logical TRUE TRUE, FALSE -
max_depth integer 20 [0,∞)[0, \infty)
max_runtime_secs numeric 0 [0,∞)[0, \infty)
min_rows numeric 1 [1,∞)[1, \infty)
min_split_improvement numeric 1e-05 [0,∞)[0, \infty)
mtries integer -1 [1,∞)[1, \infty)
nbins integer 20 [1,∞)[1, \infty)
nbins_cats integer 1024 [1,∞)[1, \infty)
nbins_top_level integer 1024 [1,∞)[1, \infty)
ntrees integer 50 [1,∞)[1, \infty)
sample_rate numeric 0.632 [0,1][0, 1]
score_each_iteration logical FALSE TRUE, FALSE -
score_tree_interval integer 0 [0,∞)[0, \infty)
seed integer -1 (−∞,∞)(-\infty, \infty)
stopping_metric character AUTO AUTO, deviance, MSE, RMSE, MAE, RMSLE -
stopping_rounds integer 0 [0,∞)[0, \infty)
stopping_tolerance numeric 0.001 [0,∞)[0, \infty)
verbose logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrH2ORandomForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrH2ORandomForest$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrH2ORandomForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Fryda T, LeDell E, Gill N, Aiello S, Fu A, Candel A, Click C, Kraljevic T, Nykodym T, Aboyoun P, Kurka M, Malohlava M, Poirier S, Wong W (2025). h2o: R Interface for the 'H2O' Scalable Machine Learning Platform. R package version 3.46.0.9, https://github.com/h2oai/h2o-3.

See Also

Examples

# Define the Learner
learner = lrn("regr.h2o.randomForest")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression IBk Learner

Description

Instance based algorithm: K-nearest neighbours regression. Calls RWeka::IBk() from RWeka.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.IBk")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
weight character - I, F -
K integer 1 [1,∞)[1, \infty)
E logical FALSE TRUE, FALSE -
W integer 0 [0,∞)[0, \infty)
X logical FALSE TRUE, FALSE -
A character LinearNNSearch BallTree, CoverTree, FilteredNeighbourSearch, KDTree, LinearNNSearch -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

  • weight:

    • original id: I and F

  • Reason for change: original I and F params are interdependent (I can only be TRUE when F is FALSE and vice versa). The easiest way to encode this is to combine I and F into one factor param.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrIBk

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrIBk$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrIBk$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrIBk$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrIBk$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

henrifnk

References

Aha, W D, Kibler, Dennis, Albert, K M (1991). “Instance-based learning algorithms.” Machine learning, 6(1), 37–66.

See Also

Examples

# Define the Learner
learner = lrn("regr.IBk")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression KStar Learner

Description

Instance-based regressor which differs from other instance-based learners in that it uses an entropy-based distance function. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.kstar")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
B integer 20 (−∞,∞)(-\infty, \infty)
E logical - TRUE, FALSE -
M character a a, d, m, n -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrKStar

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrKStar$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrKStar$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrKStar$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrKStar$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Cleary JG, Trigg LE (1995). “K*: An Instance-based Learner Using an Entropic Distance Measure.” In 12th International Conference on Machine Learning, 108-114.

See Also

Examples

# Define the Learner
learner = lrn("regr.kstar")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Kernlab Support Vector Machine

Description

Support Vector Regression. Calls kernlab::ksvm() from kernlab.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.ksvm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, kernlab

Parameters

Id Type Default Levels Range
scaled logical TRUE TRUE, FALSE -
type character eps-svr eps-svr, nu-svr, eps-bsvr -
kernel character rbfdot rbfdot, polydot, vanilladot, laplacedot, besseldot, anovadot -
C numeric 1 (−∞,∞)(-\infty, \infty)
nu numeric 0.2 [0,∞)[0, \infty)
epsilon numeric 0.1 (−∞,∞)(-\infty, \infty)
cache integer 40 [1,∞)[1, \infty)
tol numeric 0.001 [0,∞)[0, \infty)
shrinking logical TRUE TRUE, FALSE -
sigma numeric - [0,∞)[0, \infty)
degree integer - [1,∞)[1, \infty)
scale numeric - [0,∞)[0, \infty)
order integer - (−∞,∞)(-\infty, \infty)
offset numeric - (−∞,∞)(-\infty, \infty)
na.action untyped na.omit -
fit logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrKSVM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrKSVM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrKSVM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

mboecker

References

Karatzoglou, Alexandros, Smola, Alex, Hornik, Kurt, Zeileis, Achim (2004). “kernlab-an S4 package for kernel methods in R.” Journal of statistical software, 11(9), 1–20.

See Also

Examples

# Define the Learner
learner = lrn("regr.ksvm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Local Approximate Gaussian Process Learner

Description

Local approximate Gaussian process for regression. Calls laGP::aGP() from laGP.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.laGP")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, laGP

Parameters

Id Type Default Levels Range
start integer 6 [6,∞)[6, \infty)
end integer 50 [6,∞)[6, \infty)
d untyped NULL -
g untyped 1/10000 -
method character alc alc, alcray, efi, mspe, nn -
close integer - [0,∞)[0, \infty)
numrays integer - [0,∞)[0, \infty)
verb numeric 0 (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrLaGP

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrLaGP$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrLaGP$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

Parameters start and end tune the initial and maximum neighborhood sizes used for the local GP fit, d can fix the length-scale parameters, g configures the nugget/regularization term, and method selects the search heuristic (alc, alcray, efi, mspe, nn). Arguments close and numrays further refine the ray-based search used by method = "alcray", while verb controls the output verbosity.

Author(s)

awinterstetter

References

Gramacy, B. R (2016). “laGP: Large-Scale Spatial Modeling via Local Approximate Gaussian Processes in R.” Journal of Statistical Software, 72(1), 1–46. doi:10.18637/jss.v072.i01.

See Also

Examples

# Define the Learner (use a small neighborhood size for this task)
learner = lrn("regr.laGP", end = 15)
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

L2-Regularized Support Vector Regression Learner

Description

L2 regularized support vector regression. Calls LiblineaR::LiblineaR() from LiblineaR.

Details

Type of SVR depends on type argument:

  • type = 11 - L2-regularized L2-loss support vector regression (primal)

  • type = 12 – L2-regularized L2-loss support vector regression (dual)

  • type = 13 – L2-regularized L1-loss support vector regression (dual)

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.liblinear")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, LiblineaR

Parameters

Id Type Default Levels Range
type integer 11 [11,13][11, 13]
cost numeric 1 [0,∞)[0, \infty)
bias numeric 1 (−∞,∞)(-\infty, \infty)
svr_eps numeric NULL [0,∞)[0, \infty)
cross integer 0 [0,∞)[0, \infty)
verbose logical FALSE TRUE, FALSE -
findC logical FALSE TRUE, FALSE -
useInitC logical TRUE TRUE, FALSE -

Initial parameter values

  • svr_eps:

    • Actual default: NULL

    • Initial value: 0.001

    • Reason for change: svr_eps is type dependent and the "type" is handled by the mlr3learner. The default value is set to the default of the respective "type".

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrLiblineaR

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrLiblineaR$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrLiblineaR$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Fan, Rong-En, Chang, Kai-Wei, Hsieh, Cho-Jui, Wang, Xiang-Rui, Lin, Chih-Jen (2008). “LIBLINEAR: A library for large linear classification.” the Journal of machine Learning research, 9, 1871–1874.

See Also

Examples

# Define the Learner
learner = lrn("regr.liblinear")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression LightGBM Learner

Description

Gradient boosting algorithm. Calls lightgbm::lightgbm() from lightgbm. The list of parameters can be found here and in the documentation of lightgbm::lgb.train().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.lightgbm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, lightgbm

Parameters

Id Type Default Levels Range
objective character regression regression, regression_l1, huber, fair, poisson, quantile, mape, gamma, tweedie -
eval untyped - -
verbose integer 1 (−∞,∞)(-\infty, \infty)
record logical TRUE TRUE, FALSE -
eval_freq integer 1 [1,∞)[1, \infty)
callbacks untyped - -
reset_data logical FALSE TRUE, FALSE -
boosting character gbdt gbdt, rf, dart, goss -
linear_tree logical FALSE TRUE, FALSE -
learning_rate numeric 0.1 [0,∞)[0, \infty)
num_leaves integer 31 [1,131072][1, 131072]
tree_learner character serial serial, feature, data, voting -
num_threads integer 0 [0,∞)[0, \infty)
device_type character cpu cpu, gpu -
seed integer - (−∞,∞)(-\infty, \infty)
deterministic logical FALSE TRUE, FALSE -
data_sample_strategy character bagging bagging, goss -
force_col_wise logical FALSE TRUE, FALSE -
force_row_wise logical FALSE TRUE, FALSE -
histogram_pool_size integer -1 (−∞,∞)(-\infty, \infty)
max_depth integer -1 (−∞,∞)(-\infty, \infty)
min_data_in_leaf integer 20 [0,∞)[0, \infty)
min_sum_hessian_in_leaf numeric 0.001 [0,∞)[0, \infty)
bagging_fraction numeric 1 [0,1][0, 1]
bagging_freq integer 0 [0,∞)[0, \infty)
bagging_seed integer 3 (−∞,∞)(-\infty, \infty)
bagging_by_query logical FALSE TRUE, FALSE -
feature_fraction numeric 1 [0,1][0, 1]
feature_fraction_bynode numeric 1 [0,1][0, 1]
feature_fraction_seed integer 2 (−∞,∞)(-\infty, \infty)
extra_trees logical FALSE TRUE, FALSE -
extra_seed integer 6 (−∞,∞)(-\infty, \infty)
max_delta_step numeric 0 (−∞,∞)(-\infty, \infty)
lambda_l1 numeric 0 [0,∞)[0, \infty)
lambda_l2 numeric 0 [0,∞)[0, \infty)
linear_lambda numeric 0 [0,∞)[0, \infty)
min_gain_to_split numeric 0 [0,∞)[0, \infty)
drop_rate numeric 0.1 [0,1][0, 1]
max_drop integer 50 (−∞,∞)(-\infty, \infty)
skip_drop numeric 0.5 [0,1][0, 1]
xgboost_dart_mode logical FALSE TRUE, FALSE -
uniform_drop logical FALSE TRUE, FALSE -
drop_seed integer 4 (−∞,∞)(-\infty, \infty)
top_rate numeric 0.2 [0,1][0, 1]
other_rate numeric 0.1 [0,1][0, 1]
min_data_per_group integer 100 [1,∞)[1, \infty)
max_cat_threshold integer 32 [1,∞)[1, \infty)
cat_l2 numeric 10 [0,∞)[0, \infty)
cat_smooth numeric 10 [0,∞)[0, \infty)
max_cat_to_onehot integer 4 [1,∞)[1, \infty)
top_k integer 20 [1,∞)[1, \infty)
monotone_constraints untyped NULL -
monotone_constraints_method character basic basic, intermediate, advanced -
monotone_penalty numeric 0 [0,∞)[0, \infty)
feature_contri untyped NULL -
forcedsplits_filename untyped "" -
refit_decay_rate numeric 0.9 [0,1][0, 1]
cegb_tradeoff numeric 1 [0,∞)[0, \infty)
cegb_penalty_split numeric 0 [0,∞)[0, \infty)
cegb_penalty_feature_lazy untyped - -
cegb_penalty_feature_coupled untyped - -
path_smooth numeric 0 [0,∞)[0, \infty)
interaction_constraints untyped - -
use_quantized_grad logical TRUE TRUE, FALSE -
num_grad_quant_bins integer 4 (−∞,∞)(-\infty, \infty)
quant_train_renew_leaf logical FALSE TRUE, FALSE -
stochastic_rounding logical TRUE TRUE, FALSE -
serializable logical TRUE TRUE, FALSE -
max_bin integer 255 [2,∞)[2, \infty)
max_bin_by_feature untyped NULL -
min_data_in_bin integer 3 [1,∞)[1, \infty)
bin_construct_sample_cnt integer 200000 [1,∞)[1, \infty)
data_random_seed integer 1 (−∞,∞)(-\infty, \infty)
is_enable_sparse logical TRUE TRUE, FALSE -
enable_bundle logical TRUE TRUE, FALSE -
use_missing logical TRUE TRUE, FALSE -
zero_as_missing logical FALSE TRUE, FALSE -
feature_pre_filter logical TRUE TRUE, FALSE -
pre_partition logical FALSE TRUE, FALSE -
two_round logical FALSE TRUE, FALSE -
forcedbins_filename untyped "" -
boost_from_average logical TRUE TRUE, FALSE -
reg_sqrt logical FALSE TRUE, FALSE -
alpha numeric 0.9 [0,∞)[0, \infty)
fair_c numeric 1 [0,∞)[0, \infty)
poisson_max_delta_step numeric 0.7 [0,∞)[0, \infty)
tweedie_variance_power numeric 1.5 [1,2][1, 2]
metric_freq integer 1 [1,∞)[1, \infty)
num_machines integer 1 [1,∞)[1, \infty)
local_listen_port integer 12400 [1,∞)[1, \infty)
time_out integer 120 [1,∞)[1, \infty)
machines untyped "" -
gpu_platform_id integer -1 (−∞,∞)(-\infty, \infty)
gpu_device_id integer -1 (−∞,∞)(-\infty, \infty)
gpu_device_id_list untyped NULL -
gpu_use_dp logical FALSE TRUE, FALSE -
num_gpu integer 1 [1,∞)[1, \infty)
start_iteration_predict integer 0 (−∞,∞)(-\infty, \infty)
num_iteration_predict integer -1 (−∞,∞)(-\infty, \infty)
pred_early_stop logical FALSE TRUE, FALSE -
pred_early_stop_freq integer 10 (−∞,∞)(-\infty, \infty)
pred_early_stop_margin numeric 10 (−∞,∞)(-\infty, \infty)
num_iterations integer 100 [1,∞)[1, \infty)
early_stopping_rounds integer - [1,∞)[1, \infty)
early_stopping_min_delta numeric - [0,∞)[0, \infty)
first_metric_only logical FALSE TRUE, FALSE -

Initial parameter values

  • num_threads:

    • Actual default: 0L

    • Initital value: 1L

    • Reason for change: Prevents accidental conflicts with future.

  • verbose:

    • Actual default: 1L

    • Initial value: -1L

    • Reason for change: Prevents accidental conflicts with mlr messaging system.

Early Stopping and Validation

Early stopping can be used to find the optimal number of boosting rounds. Set early_stopping_rounds to an integer value to monitor the performance of the model on the validation set while training. For information on how to configure the validation set, see the Validation section of mlr3::Learner. The internal validation measure can be set the eval parameter which should be a list of mlr3::Measures, functions, or strings for the internal lightgbm measures. If first_metric_only = FALSE (default), the learner stops when any metric fails to improve.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrLightGBM

Active bindings

internal_valid_scores

The last observation of the validation scores for all metrics. Extracted from model$evaluation_log

internal_tuned_values

Returns the early stopped iterations if early_stopping_rounds was set during training.

validate

How to construct the internal validation data. This parameter can be either NULL, a ratio, "test", or "predefined".

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrLightGBM$new()

Method importance()

The importance scores are extracted from lbg.importance.

Usage
LearnerRegrLightGBM$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrLightGBM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

kapsner

References

Ke, Guolin, Meng, Qi, Finley, Thomas, Wang, Taifeng, Chen, Wei, Ma, Weidong, Ye, Qiwei, Liu, Tie-Yan (2017). “Lightgbm: A highly efficient gradient boosting decision tree.” Advances in neural information processing systems, 30.

See Also

Examples

# Define the Learner
learner = lrn("regr.lightgbm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Linear Regression Learner From Weka

Description

Linear Regression learner that uses the Akaike criterion for model selection and is able to deal with weighted instances. Calls RWeka::LinearRegression() RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • additional_stats:

    • original id: additional-stats

  • use_qr:

    • original id: use-qr

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.linear_regression")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
S character 0 0, 1, 2 -
C logical FALSE TRUE, FALSE -
R numeric 1e-08 (−∞,∞)(-\infty, \infty)
minimal logical FALSE TRUE, FALSE -
additional_stats logical FALSE TRUE, FALSE -
use_qr logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrLinearRegression

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrLinearRegression$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrLinearRegression$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrLinearRegression$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrLinearRegression$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("regr.linear_regression")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Linear Mixed Effects Learner

Description

Linear model with random effects. Calls lme4::lmer() from lme4.

Formula

Although most mlr3 learners don't allow to specify the formula manually, and automatically set it by valling task$formula(), this learner allows to set the formula because it's core functionality depends it. This means that it might not always use all features that are available in the task. Be aware, that this can sometimes lead to unexpected error messages, because mlr3 checks the compatibility between the learner and the task on all available features.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.lmer")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, lme4

Parameters

Id Type Default Levels Range
formula untyped - -
REML logical TRUE TRUE, FALSE -
start untyped NULL -
verbose integer 0 [0,∞)[0, \infty)
contrasts untyped NULL -
optimizer character nloptwrap Nelder_Mead, bobyqa, nlminbwrap, nloptwrap -
restart_edge logical FALSE TRUE, FALSE -
boundary.tol numeric 1e-05 [0,∞)[0, \infty)
calc.derivs logical TRUE TRUE, FALSE -
autoscale untyped NULL -
check.nobs.vs.rankZ character ignore ignore, warning, message, stop -
check.nobs.vs.nlev character stop ignore, warning, message, stop -
check.nlev.gtreq.5 character ignore ignore, warning, message, stop -
check.nlev.gtr.1 character stop ignore, warning, message, stop -
check.nobs.vs.nRE character stop ignore, warning, message, stop -
check.rankX character message+drop.cols message+drop.cols, silent.drop.cols, warn+drop.cols, stop.deficient, ignore -
check.scaleX character warning warning, stop, silent.rescale, message+rescale, warn+rescale, ignore -
check.formula.LHS character stop ignore, warning, message, stop -
check.conv.nobsmax numeric 10000 [1,∞)[1, \infty)
check.conv.nparmax numeric 10 [1,∞)[1, \infty)
check.conv.grad untyped "lme4::.makeCC(\"warning\", tol = 2e-3, relTol = NULL)" -
check.conv.singular untyped "lme4::.makeCC(action = \"message\", tol = formals(lme4::isSingular)$tol)" -
check.conv.hess untyped "lme4::.makeCC(action = \"warning\", tol = 1e-6)" -
optCtrl untyped list() -
newparams untyped NULL -
re.form untyped NULL -
random.only logical FALSE TRUE, FALSE -
allow.new.levels logical FALSE TRUE, FALSE -
na.action untyped "stats::na.pass" -

Offset

If a Task contains a column with the offset role, it is automatically incorporated during training via the offset argument in lme4::lmer(). No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrLmer

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrLmer$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrLmer$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

s-kganz

References

Bates, M D (2010). “lme4: Mixed-effects modeling with R.”

See Also

Examples

# Define the Learner and set parameter values
learner = lrn("regr.lmer", formula = mpg ~ (1 | gear))

# Define a Task
task = tsk("mtcars")

# Train the learner
learner$train(task)

print(learner$model)

Regression M5P Learner

Description

Implements base routines for generating M5 Model trees and rules. Calls RWeka::M5P() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.m5p")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
N logical - TRUE, FALSE -
U logical - TRUE, FALSE -
R logical - TRUE, FALSE -
M integer 4 (−∞,∞)(-\infty, \infty)
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
L logical - TRUE, FALSE -
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrM5P

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrM5P$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrM5P$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrM5P$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrM5P$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Quinlan RJ (1992). “Learning with Continuous Classes.” In 5th Australian Joint Conference on Artificial Intelligence, 343-348.

Wang Y, Witten IH (1997). “Induction of model trees for predicting continuous classes.” In Poster papers of the 9th European Conference on Machine Learning.

See Also

Examples

# Define the Learner
learner = lrn("regr.m5p")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression M5Rules Learner

Description

Algorithm for inducing decision lists from model trees. Calls RWeka::M5Rules() from RWeka.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.M5Rules")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
N logical FALSE TRUE, FALSE -
U logical FALSE TRUE, FALSE -
R logical FALSE TRUE, FALSE -
M integer 4 (−∞,∞)(-\infty, \infty)
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrM5Rules

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrM5Rules$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrM5Rules$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrM5Rules$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrM5Rules$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

henrifnk

References

Holmes, Geoffrey, Hall, Mark, Prank, Eibe (1999). “Generating rule sets from model trees.” In Australasian joint conference on artificial intelligence, 1–12. Springer.

See Also

Examples

# Define the Learner
learner = lrn("regr.M5Rules")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Mars Learner

Description

Multivariate Adaptive Regression Splines. Calls mda::mars() from mda.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.mars")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, mlr3extralearners, mda

Parameters

Id Type Default Levels Range
degree integer 1 [1,∞)[1, \infty)
nk integer - [1,∞)[1, \infty)
penalty numeric 2 [0,∞)[0, \infty)
thresh numeric 0.001 [0,∞)[0, \infty)
prune logical TRUE TRUE, FALSE -
trace.mars logical FALSE TRUE, FALSE -
forward.step logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrMars

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrMars$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrMars$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sumny

References

Friedman, H J (1991). “Multivariate adaptive regression splines.” The annals of statistics, 19(1), 1–67.

See Also

Examples

# Define the Learner
learner = lrn("regr.mars")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Model-based Recursive Partitioning Learner

Description

Model-based recursive partitioning algorithm. Calls partykit::mob() from mob.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.mob")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, partykit, sandwich, coin

Parameters

Id Type Default Levels Range
rhs untyped - -
fit untyped - -
offset untyped - -
cluster untyped - -
alpha numeric 0.05 [0,1][0, 1]
bonferroni logical TRUE TRUE, FALSE -
minsize integer - [1,∞)[1, \infty)
minsplit integer - [1,∞)[1, \infty)
minbucket integer - [1,∞)[1, \infty)
maxdepth integer Inf [0,∞)[0, \infty)
mtry integer Inf [0,∞)[0, \infty)
trim numeric 0.1 [0,∞)[0, \infty)
breakties logical FALSE TRUE, FALSE -
parm untyped - -
dfsplit integer - [0,∞)[0, \infty)
prune untyped - -
restart logical TRUE TRUE, FALSE -
verbose logical FALSE TRUE, FALSE -
caseweights logical TRUE TRUE, FALSE -
ytype character vector vector, matrix, data.frame -
xtype character matrix vector, matrix, data.frame -
terminal untyped "object" -
inner untyped "object" -
model logical TRUE TRUE, FALSE -
numsplit character left left, center -
catsplit character binary binary, multiway -
vcov character opg opg, info, sandwich -
ordinal character chisq chisq, max, L2 -
nrep integer 10000 [0,∞)[0, \infty)
applyfun untyped - -
cores integer NULL (−∞,∞)(-\infty, \infty)
additional untyped - -
predict_fun untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrMob

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrMob$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrMob$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sumny

References

Hothorn T, Zeileis A (2015). “partykit: A Modular Toolkit for Recursive Partytioning in R.” Journal of Machine Learning Research, 16(118), 3905-3909. http://jmlr.org/papers/v16/hothorn15a.html.

Hothorn T, Hornik K, Zeileis A (2006). “Unbiased Recursive Partitioning: A Conditional Inference Framework.” Journal of Computational and Graphical Statistics, 15(3), 651–674. doi:10.1198/106186006x133933, https://doi.org/10.1198/106186006x133933.

See Also

Examples

library(mlr3)
lm_ = function(y, x, start = NULL, weights = NULL, offset = NULL, ...) {
  lm(y ~ 1, ...)
}
learner = LearnerRegrMob$new()
learner$param_set$values$rhs = "."
learner$param_set$values$fit = lm_
learner$feature_types = c("logical", "integer", "numeric", "factor", "ordered")

predict_fun = function(object, newdata, task, .type) {
  preds = predict(object, newdata = newdata, type = "response", se.fit = TRUE)
  cbind(preds$fit, preds$se.fit)
}
learner$param_set$values$predict_fun = predict_fun
task = tsk("mtcars")
ids = partition(task)
learner$train(task, row_ids = ids$train)
learner$predict(task, row_ids = ids$test)

Generalized Additive Quantile Regression Model for Multiple Quantiles

Description

Quantile Regression with generalized additive models for fitting a learner on multiple quantiles simultaneously. Calls qgam::mqgam() from package qgam.

Form

For the form parameter, a gam formula specific to the Task is required (see example and ?mgcv::formula.gam). If no formula is provided, a fallback formula using all features in the task is used that will make the Learner behave like Linear Quantile Regression. The features specified in the formula need to be the same as columns with col_roles "feature" in the task.

Quantile

The quantiles for the Learner, i.e. qu parameter from qgam::mqgam(), is set using the values specified in learner$quantiles. The response quantile can be specified using learner$quantile_response.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.mqgam")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”, “quantiles”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, qgam

Parameters

Id Type Default Levels Range
form untyped - -
lsig numeric - (−∞,∞)(-\infty, \infty)
err numeric - [0,1][0, 1]
cluster untyped NULL -
multicore logical - TRUE, FALSE -
ncores numeric - (−∞,∞)(-\infty, \infty)
paropts untyped list() -
link untyped "identity" -
argGam untyped - -
discrete logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrMQGam

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrMQGam$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrMQGam$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

lona-k

References

Fasiolo, Matteo, Wood, N. S, Zaffran, Margaux, Nedellec, Raphael, Goude, Yannig (2017). “Fast Calibrated Additive Quantile Regression.” Journal of the American Statistical Association, 116, 1402–1412. doi:10.1080/01621459.2020.1725521.

See Also

Regression MultilayerPerceptron Learner

Description

Regressor that uses backpropagation to learn a multi-layer perceptron. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

  • G removed:

    • GUI will be opened

  • Reason for change: The parameter is removed because we don't want to launch GUI.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.multilayer_perceptron")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
L numeric 0.3 [0,1][0, 1]
M numeric 0.2 [0,1][0, 1]
N integer 500 [1,∞)[1, \infty)
V numeric 0 [0,100][0, 100]
S integer 0 [0,∞)[0, \infty)
E integer 20 [1,∞)[1, \infty)
A logical FALSE TRUE, FALSE -
B logical FALSE TRUE, FALSE -
H untyped "a" -
C logical FALSE TRUE, FALSE -
I logical FALSE TRUE, FALSE -
R logical FALSE TRUE, FALSE -
D logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrMultilayerPerceptron

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrMultilayerPerceptron$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrMultilayerPerceptron$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrMultilayerPerceptron$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrMultilayerPerceptron$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("regr.multilayer_perceptron")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Principal Component Regression Learner

Description

Principal Component Regression model. Calls pls::pcr() from pls.

Initial parameter values

  • model

    • Actual default: TRUE

    • Adjusted default: FALSE

    • Reason for change: keeps model objects compact because predictions do not require the stored model frame.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.pcr")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, pls

Parameters

Id Type Default Levels Range
ncomp integer - [1,∞)[1, \infty)
method character cppls kernelpls, widekernelpls, simpls, oscorespls, cppls, svdpc -
scale logical FALSE TRUE, FALSE -
model logical - TRUE, FALSE -
x logical FALSE TRUE, FALSE -
y logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrPcr

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrPcr$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrPcr$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

During prediction, all components ncomp used in training are used.

Author(s)

awinterstetter

References

Jolliffe, T. I (1982). “A note on the use of principal components in regression.” Journal of the Royal Statistical Society. Series C (Applied Statistics), 31(3), 300–303.

See Also

Examples

# Define the Learner
learner = lrn("regr.pcr")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression L1 and L2 Penalized Learner

Description

Penalized (L1 and L2) Logistic Regression. Calls penalized::penalized() from: penalized.

Details

The penalized and unpenalized arguments in the learner are implemented slightly differently than in penalized::penalized(). Here, there is no parameter for penalized but instead it is assumed that every variable is penalized unless stated in the unpenalized parameter.

Initial parameter values

  • trace is set to "FALSE" to disable printing output during model training.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.penalized")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, penalized

Parameters

Id Type Default Levels Range
epsilon numeric 1e-10 [0,∞)[0, \infty)
fusedl logical FALSE TRUE, FALSE -
lambda1 numeric 0 [0,∞)[0, \infty)
lambda2 numeric 0 [0,∞)[0, \infty)
maxiter integer - [1,∞)[1, \infty)
model character linear linear, poisson -
positive untyped FALSE -
standardize logical FALSE TRUE, FALSE -
startbeta untyped - -
startgamma untyped - -
steps untyped 1L -
trace logical TRUE TRUE, FALSE -
unpenalized untyped - -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrPenalized

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrPenalized$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrPenalized$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

annanzrv

References

Goeman, J J (2010). “L1 penalized estimation in the Cox proportional hazards model.” Biometrical journal, 52(1), 70–84.

See Also

Examples

# Define the Learner
learner = lrn("regr.penalized")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Partial Least Squares Regression

Description

Partial least squares regression. Calls pls::plsr() from pls.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.plsr")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, pls

Parameters

Id Type Default Levels Range
ncomp integer - [1,∞)[1, \infty)
method character kernelpls kernelpls, widekernelpls, simpls, oscorespls -
scale logical TRUE TRUE, FALSE -
center logical TRUE TRUE, FALSE -
validation character none none, CV, LOO -
model logical TRUE TRUE, FALSE -
x logical FALSE TRUE, FALSE -
y logical FALSE TRUE, FALSE -
orthScores logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrPlsr

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrPlsr$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrPlsr$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

During prediction, all components ncomp used in training are used.

Author(s)

awinterstetter

References

Mevik, Bjorn-Helge, Wehrens, Ron (2007). “The pls Package: Principal Component and Partial Least Squares Regression in R.” Journal of Statistical Software, 18(2), 1–24. doi:10.18637/jss.v018.i02.

See Also

Examples

# Define the Learner
learner = lrn("regr.plsr")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Polynomial Regression Learner

Description

Polynomial regression. Calls polyreg::polyFit() from polyreg.

Initial parameter values

  • deg: We have set this to 2, pretty arbitrarily.

  • noisy: We have set this to FALSE, to get no output on the console.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.polyFit")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, polyreg

Parameters

Id Type Default Levels Range
deg integer - [0,∞)[0, \infty)
maxInteractDeg integer - [0,∞)[0, \infty)
return_xy logical FALSE TRUE, FALSE -
returnPoly logical FALSE TRUE, FALSE -
noisy logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrPolyFit

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrPoly$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrPoly$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

See Also

Regression Priority Lasso Learner

Description

Patient outcome prediction based on multi-omics data taking practitioners’ preferences into account. Calls prioritylasso::prioritylasso() from prioritylasso.

Initial parameter values

  • family is set to "gaussian" and cannot be changed

  • type.measure set to "mse" (cross-validation measure)

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.priority_lasso")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, prioritylasso

Parameters

Id Type Default Levels Range
blocks untyped - -
max.coef untyped NULL -
block1.penalization logical TRUE TRUE, FALSE -
lambda.type character lambda.min lambda.min, lambda.1se -
standardize logical TRUE TRUE, FALSE -
nfolds integer 5 [1,∞)[1, \infty)
foldid untyped NULL -
cvoffset logical FALSE TRUE, FALSE -
cvoffsetnfolds integer 10 [1,∞)[1, \infty)
handle.missingtestdata character - none, omit.prediction, set.zero, impute.block -
include.allintercepts logical FALSE TRUE, FALSE -
use.blocks untyped "all" -
alignment character lambda lambda, fraction -
alpha numeric 1 [0,1][0, 1]
big numeric 9.9e+35 (−∞,∞)(-\infty, \infty)
devmax numeric 0.999 [0,1][0, 1]
dfmax integer - [0,∞)[0, \infty)
eps numeric 1e-06 [0,1][0, 1]
epsnr numeric 1e-08 [0,1][0, 1]
exclude untyped - -
exmx numeric 250 (−∞,∞)(-\infty, \infty)
fdev numeric 1e-05 [0,1][0, 1]
gamma untyped - -
grouped logical TRUE TRUE, FALSE -
intercept logical TRUE TRUE, FALSE -
keep logical FALSE TRUE, FALSE -
lambda untyped - -
lambda.min.ratio numeric - [0,1][0, 1]
lower.limits untyped -Inf -
maxit integer 100000 [1,∞)[1, \infty)
mnlam integer 5 [1,∞)[1, \infty)
mxit integer 100 [1,∞)[1, \infty)
mxitnr integer 25 [1,∞)[1, \infty)
nlambda integer 100 [1,∞)[1, \infty)
offset untyped NULL -
parallel logical FALSE TRUE, FALSE -
penalty.factor untyped - -
pmax integer - [0,∞)[0, \infty)
pmin numeric 1e-09 [0,1][0, 1]
prec numeric 1e-10 (−∞,∞)(-\infty, \infty)
standardize.response logical FALSE TRUE, FALSE -
thresh numeric 1e-07 [0,∞)[0, \infty)
trace.it integer 0 [0,1][0, 1]
type.gaussian character - covariance, naive -
type.logistic character Newton Newton, modified.Newton -
type.multinomial character ungrouped ungrouped, grouped -
upper.limits untyped Inf -
scale.y logical FALSE TRUE, FALSE -
return.x logical TRUE TRUE, FALSE -
predict.gamma numeric gamma.1se (−∞,∞)(-\infty, \infty)
relax logical FALSE TRUE, FALSE -
s numeric lambda.1se [0,1][0, 1]

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrPriorityLasso

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrPriorityLasso$new()

Method selected_features()

Selected features when coef is positive

Usage
LearnerRegrPriorityLasso$selected_features()
Returns

character().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrPriorityLasso$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

HarutyunyanLiana

References

Simon K, Vindi J, Roman H, Tobias H, Anne-Laure B (2018). “Priority-Lasso: a simple hierarchical approach to the prediction of clinical outcome using multi-omics data.” BMC Bioinformatics, 19. doi:10.1186/s12859-018-2344-6.

See Also

Examples

# Define the Learner and set parameter values
learner = lrn("regr.priority_lasso",
  blocks = list(bp1 = 1:4, bp2 = 5:9, bp3 = 10:28, bp4 = 29:1028))
print(learner)

# Define a Task
task = mlr3::as_task_regr(prioritylasso::pl_data, target = "pl_out")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

 # Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Generalized Additive Quantile Regression Model

Description

Quantile Regression with generalized additive models. Calls qgam::qgam() from package qgam.

Form

For the form parameter, a gam formula specific to the Task is required (see example and ?mgcv::formula.gam). If no formula is provided, a fallback formula using all features in the task is used that will make the Learner behave like Linear Quantile Regression. The features specified in the formula need to be the same as columns with col_roles "feature" in the task.

Quantile

The quantile for the Learner, i.e. qu parameter from qgam::qgam(), is set using the value specified in learner$quantiles.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.qgam")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “se”, “quantiles”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, qgam

Parameters

Id Type Default Levels Range
form untyped - -
lsig numeric - (−∞,∞)(-\infty, \infty)
err numeric - [0,1][0, 1]
cluster untyped NULL -
multicore logical - TRUE, FALSE -
ncores numeric - (−∞,∞)(-\infty, \infty)
paropts untyped list() -
link untyped "identity" -
argGam untyped - -
discrete logical FALSE TRUE, FALSE -
block.size integer 1000 (−∞,∞)(-\infty, \infty)
unconditional logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrQGam

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrQGam$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrQGam$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

lona-k

References

Fasiolo, Matteo, Wood, N. S, Zaffran, Margaux, Nedellec, Raphael, Goude, Yannig (2017). “Fast Calibrated Additive Quantile Regression.” Journal of the American Statistical Association, 116, 1402–1412. doi:10.1080/01621459.2020.1725521.

See Also

Regression Random Forest Learner from Weka

Description

Class for constructing a forest of random trees. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • store_out_of_bag_predictions:

    • original id: store-out-of-bag-predictions

  • output_out_of_bag_complexity_statistics:

    • original id: output-out-of-bag-complexity-statistics

  • num_slots:

    • original id: num-slots

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

  • attribute-importance removed:

    • Compute and output attribute importance (mean impurity decrease method)

  • Reason for change: The parameter is removed because it's unclear how to actually use it.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.random_forest_weka")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
P numeric 100 [0,100][0, 100]
O logical FALSE TRUE, FALSE -
store_out_of_bag_predictions logical FALSE TRUE, FALSE -
output_out_of_bag_complexity_statistics logical FALSE TRUE, FALSE -
print logical FALSE TRUE, FALSE -
I integer 100 [1,∞)[1, \infty)
num_slots integer 1 (−∞,∞)(-\infty, \infty)
K integer 0 (−∞,∞)(-\infty, \infty)
M integer 1 [1,∞)[1, \infty)
V numeric 0.001 (−∞,∞)(-\infty, \infty)
S integer 1 (−∞,∞)(-\infty, \infty)
depth integer 0 [0,∞)[0, \infty)
N integer 0 (−∞,∞)(-\infty, \infty)
U logical FALSE TRUE, FALSE -
B logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrRandomForestWeka

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrRandomForestWeka$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrRandomForestWeka$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrRandomForestWeka$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrRandomForestWeka$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Breiman, Leo (2001). “Random Forests.” Machine Learning, 45(1), 5–32. ISSN 1573-0565, doi:10.1023/A:1010933404324.

See Also

Examples

# Define the Learner
learner = lrn("regr.random_forest_weka")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Random Tree Learner

Description

Tree that considers K randomly chosen attributes at each node. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.random_tree")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
K integer 0 [0,∞)[0, \infty)
M integer 1 [1,∞)[1, \infty)
V numeric 0.001 (−∞,∞)(-\infty, \infty)
S integer 1 (−∞,∞)(-\infty, \infty)
depth integer 0 [0,∞)[0, \infty)
N integer 0 [0,∞)[0, \infty)
U logical FALSE TRUE, FALSE -
B logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrRandomTree

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrRandomTree$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrRandomTree$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrRandomTree$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrRandomTree$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("regr.random_tree")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Random Forest Learner

Description

Random forest for regression. Calls randomForest::randomForest() from randomForest.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.randomForest")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, randomForest

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
mtry integer - [1,∞)[1, \infty)
replace logical TRUE TRUE, FALSE -
strata untyped - -
sampsize untyped - -
nodesize integer 5 [1,∞)[1, \infty)
maxnodes integer - [1,∞)[1, \infty)
importance character FALSE mse, nudepurity, none -
localImp logical FALSE TRUE, FALSE -
proximity logical FALSE TRUE, FALSE -
oob.prox logical - TRUE, FALSE -
norm.votes logical TRUE TRUE, FALSE -
do.trace logical FALSE TRUE, FALSE -
keep.forest logical TRUE TRUE, FALSE -
keep.inbag logical FALSE TRUE, FALSE -
predict.all logical FALSE TRUE, FALSE -
nodes logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrRandomForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrRandomForest$new()

Method importance()

The importance scores are extracted from the slot importance. Parameter 'importance' must be set to either "mse" or "nodepurity".

Usage
LearnerRegrRandomForest$importance()
Returns

Named numeric().

Method oob_error()

OOB errors are extracted from the model slot mse.

Usage
LearnerRegrRandomForest$oob_error()
Returns

numeric(1).

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrRandomForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

pat-s

References

Breiman, Leo (2001). “Random Forests.” Machine Learning, 45(1), 5–32. ISSN 1573-0565, doi:10.1023/A:1010933404324.

See Also

Examples

# Define the Learner
learner = lrn("regr.randomForest", importance = "mse")
print(learner)

# Define a Task
task = tsk("mtcars")
# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Decision Tree Learner

Description

Fast decision tree learner. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.reptree")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
M integer 2 (−∞,∞)(-\infty, \infty)
V numeric 0.001 (−∞,∞)(-\infty, \infty)
N integer 3 (−∞,∞)(-\infty, \infty)
S integer 1 (−∞,∞)(-\infty, \infty)
P logical - TRUE, FALSE -
L integer -1 (−∞,∞)(-\infty, \infty)
I integer 0 (−∞,∞)(-\infty, \infty)
R logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrREPTree

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrREPTree$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrREPTree$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrREPTree$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrREPTree$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("regr.reptree")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Random Forest SRC Learner

Description

Random forest for regression. Calls randomForestSRC::rfsrc() from randomForestSRC.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.rfsrc")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, randomForestSRC

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
mtry integer - [1,∞)[1, \infty)
mtry.ratio numeric - [0,1][0, 1]
nodesize integer 15 [1,∞)[1, \infty)
nodedepth integer - [1,∞)[1, \infty)
splitrule character mse mse, quantile.regr, la.quantile.regr -
nsplit integer 10 [0,∞)[0, \infty)
importance character FALSE FALSE, TRUE, none, permute, random, anti -
block.size integer 10 [1,∞)[1, \infty)
bootstrap character by.root by.root, by.node, none, by.user -
samptype character swor swor, swr -
samp untyped - -
membership logical FALSE TRUE, FALSE -
sampsize untyped - -
sampsize.ratio numeric - [0,1][0, 1]
na.action character na.omit na.omit, na.impute -
nimpute integer 1 [1,∞)[1, \infty)
proximity character FALSE FALSE, TRUE, inbag, oob, all -
distance character FALSE FALSE, TRUE, inbag, oob, all -
forest.wt character FALSE FALSE, TRUE, inbag, oob, all -
xvar.wt untyped - -
split.wt untyped - -
forest logical TRUE TRUE, FALSE -
var.used character FALSE FALSE, all.trees -
split.depth character FALSE FALSE, all.trees, by.tree -
seed integer - (−∞,−1](-\infty, -1]
do.trace logical FALSE TRUE, FALSE -
get.tree untyped - -
outcome character train train, test -
ptn.count integer 0 [0,∞)[0, \infty)
cores integer 1 [1,∞)[1, \infty)
save.memory logical FALSE TRUE, FALSE -
perf.type character - none -
case.depth logical FALSE TRUE, FALSE -
marginal.xvar untyped NULL -

Custom mlr3 parameters

  • mtry: This hyperparameter can alternatively be set via the added hyperparameter mtry.ratio as mtry = max(ceiling(mtry.ratio * n_features), 1). Note that mtry and mtry.ratio are mutually exclusive.

  • sampsize: This hyperparameter can alternatively be set via the added hyperparameter sampsize.ratio as sampsize = max(ceiling(sampsize.ratio * n_obs), 1). Note that sampsize and sampsize.ratio are mutually exclusive.

  • cores: This value is set as the option rf.cores during training and is set to 1 by default.

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrRandomForestSRC

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrRandomForestSRC$new()

Method importance()

The importance scores are extracted from the model slot importance.

Usage
LearnerRegrRandomForestSRC$importance()
Returns

Named numeric().

Method selected_features()

Selected features are extracted from the model slot var.used.

Note: Due to a known issue in randomForestSRC, enabling var.used = "all.trees" causes prediction to fail. Therefore, this setting should be used exclusively for feature selection purposes and not when prediction is required.

Usage
LearnerRegrRandomForestSRC$selected_features()
Returns

character().

Method oob_error()

OOB error extracted from the model slot err.rate.

Usage
LearnerRegrRandomForestSRC$oob_error()
Returns

numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrRandomForestSRC$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Breiman, Leo (2001). “Random Forests.” Machine Learning, 45(1), 5–32. ISSN 1573-0565, doi:10.1023/A:1010933404324.

See Also

Examples

# Define the Learner
learner = lrn("regr.rfsrc", importance = "TRUE")
print(learner)

# Define a Task
task = tsk("mtcars")
# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Regularized Random Forest Learner

Description

Regularized random forest for regression. Calls RRF::RRF() from RRF.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.RRF")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3extralearners, RRF

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
mtry integer - [1,∞)[1, \infty)
nodesize integer - [1,∞)[1, \infty)
replace logical TRUE TRUE, FALSE -
flagReg integer 1 [0,∞)[0, \infty)
coefReg numeric 0.8 (−∞,∞)(-\infty, \infty)
feaIni untyped - -
corr.bias logical FALSE TRUE, FALSE -
maxnodes integer - [1,∞)[1, \infty)
importance logical FALSE TRUE, FALSE -
localImp logical FALSE TRUE, FALSE -
nPerm integer 1 [1,∞)[1, \infty)
proximity logical FALSE TRUE, FALSE -
oob.prox logical FALSE TRUE, FALSE -
do.trace logical FALSE TRUE, FALSE -
keep.inbag logical FALSE TRUE, FALSE -
keep.forest logical TRUE TRUE, FALSE -
strata untyped - -
sampsize untyped - -
predict.all logical FALSE TRUE, FALSE -
nodes logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrRRF

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrRRF$new()

Method importance()

The importance scores are extracted from the slot importance.

Usage
LearnerRegrRRF$importance()
Returns

Named numeric().

Method oob_error()

OOB errors are extracted from the model slot mse.

Usage
LearnerRegrRRF$oob_error()
Returns

numeric(1).

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrRRF$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

awinterstetter

References

Deng, Houtao, Runger, George (2012). “Feature selection via regularized trees.” In 2012 International Joint Conference on Neural Networks (IJCNN), 1–8. IEEE. doi:10.1109/IJCNN.2012.6252640.

Deng, Houtao, Runger, George (2013). “Gene selection with guided regularized random forest.” Pattern Recognition, 46(12), 3483–3489. doi:10.1016/j.patcog.2013.05.021.

See Also

Examples

# Define the Learner
learner = lrn("regr.RRF")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Response Surface Model Learner

Description

Fit a linear model with a response-surface component. Calls rsm::rsm() from rsm.

Custom mlr3 parameters

  • modelfun: This parameter controls how the formula for rsm::rsm() is created. Possible values are:

    • "FO" - first order

    • "TWI" - wo-way interactions, this is with 1st oder terms

    • "SO" - full second order

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.rsm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, rsm

Parameters

Id Type Default Levels
modelfun character - FO, TWI, SO

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrRSM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrRSM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrRSM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

sebffischer

References

Lenth, V R (2010). “Response-surface methods in R, using rsm.” Journal of Statistical Software, 32, 1–17.

See Also

Examples

# Define the Learner
learner = lrn("regr.rsm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Relevance Vector Machine Learner

Description

Bayesian version of the support vector machine. Parameters sigma, degree, scale, offset, order, length, lambda, and normalized are added to make tuning kpar easier. If kpar is provided then these new parameters are ignored. If none are provided then the default "automatic" is used for kpar. Calls kernlab::rvm() from package kernlab.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.rvm")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3extralearners, kernlab

Parameters

Id Type Default Levels Range
kernel character rbfdot rbfdot, polydot, vanilladot, tanhdot, laplacedot, besseldot, anovadot, splinedot, stringdot -
sigma numeric - (−∞,∞)(-\infty, \infty)
degree numeric - (−∞,∞)(-\infty, \infty)
scale numeric - (−∞,∞)(-\infty, \infty)
offset numeric - (−∞,∞)(-\infty, \infty)
order numeric - (−∞,∞)(-\infty, \infty)
length integer - [0,∞)[0, \infty)
lambda numeric - (−∞,∞)(-\infty, \infty)
normalized logical - TRUE, FALSE -
kpar untyped "automatic" -
alpha untyped 5 -
var numeric 0.1 [0.001,∞)[0.001, \infty)
var.fix logical FALSE TRUE, FALSE -
iterations integer 100 [0,∞)[0, \infty)
tol numeric 2.220446e-16 [0,∞)[0, \infty)
minmaxdiff numeric 0.001 [0,∞)[0, \infty)
verbosity logical FALSE TRUE, FALSE -
fit logical TRUE TRUE, FALSE -
na.action untyped na.omit -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrRVM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrRVM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrRVM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Karatzoglou, Alexandros, Smola, Alex, Hornik, Kurt, Zeileis, Achim (2004). “kernlab-an S4 package for kernel methods in R.” Journal of statistical software, 11(9), 1–20.

See Also

Examples

# Define the Learner
learner = lrn("regr.rvm")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Stochastic Gradient Descent Learner

Description

Stochastic Gradient Descent for learning various linear models. Calls RWeka::make_Weka_classifier() from RWeka.

Initial parameter values

  • F:

    • Has only 3 out of 5 original loss functions: 2 = squared loss (regression), 3 = epsilon insensitive loss (regression) and 4 = Huber loss (regression) with 2 (squared loss) being the new default

    • Reason for change: this learner should only contain loss functions appropriate for regression tasks

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.sgd")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
F character 2 2, 3, 4 -
L numeric 0.01 (−∞,∞)(-\infty, \infty)
R numeric 1e-04 (−∞,∞)(-\infty, \infty)
E integer 500 (−∞,∞)(-\infty, \infty)
C numeric 0.001 (−∞,∞)(-\infty, \infty)
N logical - TRUE, FALSE -
M logical - TRUE, FALSE -
S integer 1 (−∞,∞)(-\infty, \infty)
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrSGD

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrSGD$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrSGD$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrSGD$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrSGD$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("regr.sgd")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression SimpleLinearRegression Learner from Weka

Description

Simple linear regression model that picks the attribute that results in the lowest squared error. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • additional_stats:

    • original id: additional-stats

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.simple_linear_regression")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “integer”, “numeric”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
additional_stats logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrSimpleLinearRegression

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrSimpleLinearRegression$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrSimpleLinearRegression$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrSimpleLinearRegression$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrSimpleLinearRegression$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

See Also

Examples

# Define the Learner
learner = lrn("regr.simple_linear_regression")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Regression Support Vector Machine Learner

Description

Support Vector Machine for regression. Calls RWeka::make_Weka_classifier() from RWeka.

Custom mlr3 parameters

  • output_debug_info:

    • original id: output-debug-info

  • do_not_check_capabilities:

    • original id: do-not-check-capabilities

  • num_decimal_places:

    • original id: num-decimal-places

  • batch_size:

    • original id: batch-size

  • T_improved:

    • original id: T

  • V_improved:

    • original id: V

  • P_improved:

    • original id: P

  • L_improved:

    • original id: L (duplicated L for when I is set to RegSMOImproved)

  • W_improved:

    • original id: W

  • C_poly:

    • original id: C

  • E_poly:

    • original id: E

  • L_poly:

    • original id: L (duplicated L for when K is set to PolyKernel)

  • Reason for change: This learner contains changed ids of the following control arguments since their ids contain irregular pattern

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.smo_reg")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, RWeka

Parameters

Id Type Default Levels Range
subset untyped - -
na.action untyped - -
C numeric 1 (−∞,∞)(-\infty, \infty)
N character 0 0, 1, 2 -
I character RegSMOImproved RegSMO, RegSMOImproved -
K character PolyKernel NormalizedPolyKernel, PolyKernel, Puk, RBFKernel, StringKernel -
T_improved numeric 0.001 (−∞,∞)(-\infty, \infty)
V_improved logical TRUE TRUE, FALSE -
P_improved numeric 1e-12 (−∞,∞)(-\infty, \infty)
L_improved numeric 0.001 (−∞,∞)(-\infty, \infty)
W_improved integer 1 (−∞,∞)(-\infty, \infty)
C_poly integer 250007 (−∞,∞)(-\infty, \infty)
E_poly numeric 1 (−∞,∞)(-\infty, \infty)
L_poly logical FALSE TRUE, FALSE -
output_debug_info logical FALSE TRUE, FALSE -
do_not_check_capabilities logical FALSE TRUE, FALSE -
num_decimal_places integer 2 [1,∞)[1, \infty)
batch_size integer 100 [1,∞)[1, \infty)
options untyped NULL -

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrSMOreg

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrSMOreg$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrSMOreg$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrSMOreg$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrSMOreg$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

damirpolat

References

Shevade S, Keerthi S, Bhattacharyya C, Murthy K (1999). “Improvements to the SMO Algorithm for SVM Regression.” In IEEE Transactions on Neural Networks.

See Also

Examples

# Define the Learner
learner = lrn("regr.smo_reg")
print(learner)

# Define a Task
task = tsk("mtcars")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

TabPFN Regression Learner

Description

Foundation model for tabular data. Uses reticulate to interface with the tabpfn Python package.

Installation

This learner relies on reticulate to handle Python dependencies. It is not necessary to install any Python package manually in advance or specify a Python environment via reticulate::use_python(), reticulate::use_virtualenv(), reticulate::use_condaenv(), or reticulate::use_miniconda(). By calling ⁠$train()⁠ or ⁠$predict()⁠, the required Python packages (tapfn, torch, etc.) will be installed automatically, if not already. Reticulate will then configure and initialize an ephemeral environment satisfying those requirements, unless an existing environment (e.g., "r-reticulate") in reticulate's Order of Discovery contains all the necessary packages.

You may also manually install tabpfn into a Python environment following the official installation guide and specify the environment via ⁠reticulate::use_*()⁠ before calling ⁠$train()⁠ or ⁠$predict()⁠. Note that the GPU version of PyTorch cannot be loaded by reticulate::use_condaenv() from a conda environment. To use a conda environment, please install the CPU version of PyTorch.

Saving a Learner

In order to save a LearnerRegrTabPFN for later usage, it is necessary to call the ⁠$marshal()⁠ method on the Learner before writing it to disk, as the object will otherwise not be saved correctly. After loading a marshaled LearnerRegrTabPFN into R again, you then need to call ⁠$unmarshal()⁠ to transform it into a useable state.

Initial parameter values

  • n_jobs is initialized to 1 to avoid threading conflicts with future.

Custom mlr3 parameters

  • output_type corresponds to the same argument of the .predict() method of the TabPFNRegressor class, but only supports the options "mean", "median" and "mode". The point predictions are stored as ⁠$response⁠ of the prediction object.

  • categorical_feature_indices uses R indexing instead of zero-based Python indexing.

  • device must be a string. If set to "auto", the behavior is the same as original. Otherwise, the string is passed as argument to torch.device() to create a device.

  • inference_precision must be "auto", "autocast", or a torch.dtype string, e.g., "torch.float32", "torch.float64", etc. Non-float dtypes are not supported.

  • inference_config is currently not supported.

  • random_state accepts either an integer or the special value "None" which corresponds to None in Python. Following the original Python implementation, the default random_state is 0.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("regr.tabpfn")

Meta Information

  • Task type: “regr”

  • Predict Types: “response”, “quantiles”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, reticulate

Parameters

Id Type Default Levels Range
output_type character mean mean, median, mode -
n_estimators integer 4 [1,∞)[1, \infty)
categorical_features_indices untyped - -
softmax_temperature numeric 0.9 [0,∞)[0, \infty)
average_before_softmax logical FALSE TRUE, FALSE -
model_path untyped "auto" -
device untyped "auto" -
ignore_pretraining_limits logical FALSE TRUE, FALSE -
inference_precision character auto auto, autocast, torch.float32, torch.float, torch.float64, torch.double, torch.float16, torch.half, torch.bfloat16 -
fit_mode character fit_preprocessors low_memory, fit_preprocessors, fit_with_cache -
memory_saving_mode untyped "auto" -
random_state integer 0 (−∞,∞)(-\infty, \infty)
n_jobs integer - [1,∞)[1, \infty)

Super classes

mlr3::Learner -> mlr3::LearnerRegr -> LearnerRegrTabPFN

Active bindings

marshaled

(logical(1))
Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerRegrTabPFN$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerRegrTabPFN$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerRegrTabPFN$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerRegrTabPFN$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

b-zhou

References

Hollmann, Noah, Müller, Samuel, Purucker, Lennart, Krishnakumar, Arjun, Körfer, Max, Hoo, Bin S, Schirrmeister, Tibor R, Hutter, Frank (2025). “Accurate predictions on small data with a tabular foundation model.” Nature. doi:10.1038/s41586-024-08328-6, https://www.nature.com/articles/s41586-024-08328-6.

Hollmann, Noah, Müller, Samuel, Eggensperger, Katharina, Hutter, Frank (2023). “TabPFN: A transformer that solves small tabular classification problems in a second.” In International Conference on Learning Representations 2023.

See Also

Accelerated Oblique Random Survival Forest Learner

Description

Accelerated oblique random survival forest. Calls aorsf::orsf() from aorsf. Note that although the learner has the property "missing" and it can in principle deal with missing values, the behaviour has to be configured using the parameter na_action.

Initial parameter values

  • n_thread: This parameter is initialized to 1 (default is 0) to avoid conflicts with the mlr3 parallelization.

  • mtry:

    • This hyperparameter can alternatively be set via the added hyperparameter mtry_ratio as mtry = max(ceiling(mtry_ratio * n_features), 1). Note that mtry and mtry_ratio are mutually exclusive.

Prediction types

This learner returns three prediction types:

  1. distr: a survival matrix in two dimensions, where observations are represented in rows and (unique event) time points in columns. Calculated using the internal predict.ObliqueForest() function.

  2. response: the restricted mean survival time of each test observation, derived from the survival matrix prediction (distr).

  3. crank: the expected mortality using mlr3proba::surv_return().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.aorsf")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “response”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, aorsf

Parameters

Id Type Default Levels Range
n_tree integer 500 [1,∞)[1, \infty)
n_split integer 5 [1,∞)[1, \infty)
n_retry integer 3 [0,∞)[0, \infty)
n_thread integer - [0,∞)[0, \infty)
pred_aggregate logical TRUE TRUE, FALSE -
pred_simplify logical FALSE TRUE, FALSE -
oobag logical FALSE TRUE, FALSE -
mtry integer NULL [1,∞)[1, \infty)
mtry_ratio numeric - [0,1][0, 1]
sample_with_replacement logical TRUE TRUE, FALSE -
sample_fraction numeric 0.632 [0,1][0, 1]
control_type character fast fast, cph, net, custom -
control_custom_fun untyped - -
split_rule character logrank logrank, cstat -
control_fast_do_scale logical FALSE TRUE, FALSE -
control_fast_ties character efron efron, breslow -
control_cph_ties character efron efron, breslow -
control_cph_eps numeric 1e-09 [0,∞)[0, \infty)
control_cph_iter_max integer 20 [1,∞)[1, \infty)
control_net_alpha numeric 0.5 (−∞,∞)(-\infty, \infty)
control_net_df_target integer NULL [1,∞)[1, \infty)
leaf_min_events integer 1 [1,∞)[1, \infty)
leaf_min_obs integer 5 [1,∞)[1, \infty)
split_min_events integer 5 [1,∞)[1, \infty)
split_min_obs integer 10 [1,∞)[1, \infty)
split_min_stat numeric NULL [0,∞)[0, \infty)
oobag_pred_type character risk none, surv, risk, chf, mort -
importance character anova none, anova, negate, permute -
importance_max_pvalue numeric 0.01 [1e−04,0.9999][1e-04, 0.9999]
tree_seeds integer NULL [1,∞)[1, \infty)
oobag_pred_horizon numeric NULL [0,∞)[0, \infty)
oobag_eval_every integer NULL [1,∞)[1, \infty)
oobag_fun untyped NULL -
attach_data logical TRUE TRUE, FALSE -
verbose_progress logical FALSE TRUE, FALSE -
na_action character fail fail, impute_meanmode -

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvAorsf

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvAorsf$new()

Method oob_error()

OOB concordance error extracted from the model slot eval_oobag$stat_values

Usage
LearnerSurvAorsf$oob_error()
Returns

numeric().

Method importance()

The importance scores are extracted from the model.

Usage
LearnerSurvAorsf$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvAorsf$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bcjaeger

References

Jaeger BC, Long DL, Long DM, Sims M, Szychowski JM, Min Y, Mcclure LA, Howard G, Simon N (2019). “Oblique random survival forests.” The Annals of Applied Statistics, 13(3). doi:10.1214/19-aoas1261.

Jaeger BC, Welden S, Lenoir K, Speiser JL, Segar MW, Pandey A, Pajewski NM (2023). “Accelerated and interpretable oblique random survival forests.” Journal of Computational and Graphical Statistics, 1–16. doi:10.1080/10618600.2023.2231048.

See Also

Examples

# Define the Learner
learner = lrn("surv.aorsf")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Survival Bayesian Additive Regression Trees Learner

Description

Fits a Bayesian Additive Regression Trees (BART) learner to right-censored survival data. Calls BART::mc.surv.bart() from BART.

Prediction types

This learner returns two prediction types:

  1. distr: a 3d survival array with observations as 1st dimension, time points as 2nd and the posterior draws as 3rd dimension. Calculated using the internal predict.survbart() function.

  2. crank: the expected mortality using mlr3proba::surv_return(). The parameter which.curve decides which posterior draw (3rd dimension) will be used for the calculation of the expected mortality. Note that the median posterior is by default used for the calculation of survival measures that require a distr prediction, see more info on PredictionSurv.

Initial parameter values

  • mc.cores is initialized to 1 to avoid threading conflicts with future.

Custom mlr3 parameters

  • quiet allows to suppress messages generated by the wrapped C++ code. Is initialized to TRUE.

  • importance allows to choose the type of importance. Default is count, see documentation of method ⁠$importance()⁠ for more details.

  • which.curve allows to choose which posterior draw will be used for the calculation of the crank prediction. If between (0,1) it is taken as the quantile of the curves otherwise if greater than 1 it is taken as the curve index, can also be 'mean'. By default the median posterior is used, i.e. which.curve is 0.5.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.bart")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3proba, BART

Parameters

Id Type Default Levels Range
K numeric NULL [1,∞)[1, \infty)
events untyped NULL -
ztimes untyped NULL -
zdelta untyped NULL -
sparse logical FALSE TRUE, FALSE -
theta numeric 0 (−∞,∞)(-\infty, \infty)
omega numeric 1 (−∞,∞)(-\infty, \infty)
a numeric 0.5 [0.5,1][0.5, 1]
b numeric 1 (−∞,∞)(-\infty, \infty)
augment logical FALSE TRUE, FALSE -
rho numeric NULL (−∞,∞)(-\infty, \infty)
usequants logical FALSE TRUE, FALSE -
rm.const logical TRUE TRUE, FALSE -
type character pbart pbart, lbart -
ntype integer - [1,3][1, 3]
k numeric 2 [0,∞)[0, \infty)
power numeric 2 [0,∞)[0, \infty)
base numeric 0.95 [0,1][0, 1]
offset numeric NULL (−∞,∞)(-\infty, \infty)
ntree integer 50 [1,∞)[1, \infty)
numcut integer 100 [1,∞)[1, \infty)
ndpost integer 1000 [1,∞)[1, \infty)
nskip integer 250 [0,∞)[0, \infty)
keepevery integer 10 [1,∞)[1, \infty)
printevery integer 100 [1,∞)[1, \infty)
seed integer 99 (−∞,∞)(-\infty, \infty)
mc.cores integer 2 [1,∞)[1, \infty)
nice integer 19 [0,19][0, 19]
openmp logical TRUE TRUE, FALSE -
quiet logical TRUE TRUE, FALSE -
importance character count count, prob -
which.curve numeric - [0,∞)[0, \infty)

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvLearnerSurvBART

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvLearnerSurvBART$new()

Method importance()

Two types of importance scores are supported based on the value of the parameter importance:

  1. prob: The mean selection probability of each feature in the trees, extracted from the slot varprob.mean. If sparse = FALSE (default), this is a fixed constant. Recommended to use this option when sparse = TRUE.

  2. count: The mean observed count of each feature in the trees (average number of times the feature was used in a tree decision rule across all posterior draws), extracted from the slot varcount.mean. This is the default importance scores.

In both cases, higher values signify more important variables.

Usage
LearnerSurvLearnerSurvBART$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvLearnerSurvBART$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Sparapani, Rodney, Spanbauer, Charles, McCulloch, Robert (2021). “Nonparametric machine learning and efficient computation with bayesian additive regression trees: the BART R package.” Journal of Statistical Software, 97, 1–66.

Chipman, A H, George, I E, McCulloch, E R (2010). “BART: Bayesian additive regression trees.” The Annals of Applied Statistics, 4(1), 266–298.

See Also

Examples

lrn("surv.bart")

Gradient Boosting with Regression Trees Survival Learner

Description

Gradient boosting with regression trees for survival analysis. Calls mboost::blackboost() from mboost.

Details

distr prediction made by mboost::survFit().

Prediction types

This learner returns two to three prediction types:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using mboost::predict.blackboost(). If the family parameter is not "coxph", -lp is returned, since non-coxph families represent AFT-style distributions where lower lp values indicate higher risk.

  2. crank: same as lp.

  3. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using mboost::survFit(). This prediction type is present only when the family distribution parameter is equal to "coxph" (default). By default the Breslow estimator is used for computing the baseline hazard.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.blackboost")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”

  • Feature Types: “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, mboost

Parameters

Id Type Default Levels Range
family character coxph coxph, weibull, loglog, lognormal, gehan, cindex, custom -
custom.family untyped - -
nuirange untyped c(0, 100) -
center logical TRUE TRUE, FALSE -
mstop integer 100 [0,∞)[0, \infty)
nu numeric 0.1 [0,1][0, 1]
risk character - inbag, oobag, none -
stopintern logical FALSE TRUE, FALSE -
trace logical FALSE TRUE, FALSE -
oobweights untyped - -
teststat character quadratic quadratic, maximum -
splitstat character quadratic quadratic, maximum -
splittest logical FALSE TRUE, FALSE -
testtype character Bonferroni Bonferroni, MonteCarlo, Univariate, Teststatistic -
maxpts integer 25000 [1,∞)[1, \infty)
abseps numeric 0.001 (−∞,∞)(-\infty, \infty)
releps numeric 0 (−∞,∞)(-\infty, \infty)
nmax untyped - -
alpha numeric 0.05 [0,1][0, 1]
mincriterion numeric 0.95 [0,1][0, 1]
logmincriterion numeric -0.05129329 (−∞,0](-\infty, 0]
minsplit integer 20 [0,∞)[0, \infty)
minbucket integer 7 [0,∞)[0, \infty)
minprob numeric 0.01 [0,1][0, 1]
stump logical FALSE TRUE, FALSE -
lookahead logical FALSE TRUE, FALSE -
MIA logical FALSE TRUE, FALSE -
nresample integer 9999 [1,∞)[1, \infty)
tol numeric 1.490116e-08 [0,∞)[0, \infty)
maxsurrogate integer 0 [0,∞)[0, \infty)
mtry integer - [0,∞)[0, \infty)
maxdepth integer - [0,∞)[0, \infty)
multiway logical FALSE TRUE, FALSE -
splittry integer 2 [1,∞)[1, \infty)
intersplit logical FALSE TRUE, FALSE -
majority logical FALSE TRUE, FALSE -
caseweights logical TRUE TRUE, FALSE -
sigma numeric 0.1 [0,1][0, 1]
ipcw untyped 1 -
na.action untyped stats::na.omit -

Offset

If a Task contains a column with the offset role, it is automatically incorporated via the offset argument in mboost's training function. No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvBlackBoost

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvBlackBoost$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvBlackBoost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Bühlmann, Peter, Yu, Bin (2003). “Boosting with the L 2 loss: regression and classification.” Journal of the American Statistical Association, 98(462), 324–339.

See Also

Examples

# Define the Learner
learner = lrn("surv.blackboost")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

BlockForest Survival Learner

Description

Random survival forests for blocks of clinical and omics covariate data. Calls blockForest::blockfor() from package blockForest.

In this learner, only the trained forest object (⁠$forest⁠) is retained. The optimized block-specific tuning parameters (paramvalues) and the biased OOB error estimate (biased_oob_error_donotuse) are discarded, as they are either not needed for downstream use or not reliable for performance estimation.

Prediction types

This learner returns two prediction types:

  1. distr: a survival matrix in two dimensions, where observations are represented in rows and (unique event) time points in columns. Calculated using the internal blockForest:::predict.blockForest() function.

  2. crank: the expected mortality using mlr3proba::surv_return().

Initial parameter values

  • num.threads is initialized to 1 to avoid conflicts with parallelization via future.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.blockforest")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, blockForest

Parameters

Id Type Default Levels Range
blocks untyped - -
block.method character BlockForest BlockForest, RandomBlock, BlockVarSel, VarProb, SplitWeights -
num.trees integer 2000 [1,∞)[1, \infty)
mtry untyped NULL -
nsets integer 300 [1,∞)[1, \infty)
num.trees.pre integer 1500 [1,∞)[1, \infty)
splitrule character extratrees logrank, extratrees, C, maxstat -
always.select.block integer 0 [0,1][0, 1]
importance character - none, impurity, impurity_corrected, permutation -
num.threads integer - [1,∞)[1, \infty)
seed integer NULL (−∞,∞)(-\infty, \infty)
verbose logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvBlockForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvBlockForest$new()

Method importance()

The importance scores are extracted from the model slot variable.importance.

Usage
LearnerSurvBlockForest$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvBlockForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Hornung, R., Wright, N. M (2019). “Block Forests: Random forests for blocks of clinical and omics covariate data.” BMC Bioinformatics, 20(1), 1–17. doi:10.1186/s12859-019-2942-y, https://doi.org/10.1186/s12859-019-2942-y.

See Also

Examples

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# check task's features
task$feature_names

# partition features to 2 blocks
blocks = list(bl1 = 1:3, bl2 = 4:6)

# define learner
learner = lrn("surv.blockforest", blocks = blocks,
              importance = "permutation", nsets = 10,
              num.trees = 50, num.trees.pre = 10, splitrule = "logrank")

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# feature importance
learner$importance()

# Make predictions for the test observations
pred = learner$predict(task, row_ids = ids$test)
pred

# Score the predictions
pred$score()

Survival Conditional Random Forest Learner

Description

A random forest based on conditional inference trees (partykit::ctree()). Calls partykit::cforest() from partykit.

Prediction types

This learner returns two prediction types:

  1. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using the internal partykit::predict.cforest() function.

  2. crank: the expected mortality using mlr3proba::surv_return().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.cforest")

Meta Information

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
replace logical FALSE TRUE, FALSE -
fraction numeric 0.632 [0,1][0, 1]
mtry integer - [0,∞)[0, \infty)
mtryratio numeric - [0,1][0, 1]
applyfun untyped - -
cores integer NULL (−∞,∞)(-\infty, \infty)
trace logical FALSE TRUE, FALSE -
cluster untyped - -
na.action untyped "stats::na.pass" -
scores untyped - -
teststat character quadratic quadratic, maximum -
splitstat character quadratic quadratic, maximum -
splittest logical FALSE TRUE, FALSE -
testtype character Univariate Bonferroni, MonteCarlo, Univariate, Teststatistic -
nmax untyped - -
alpha numeric 0.05 [0,1][0, 1]
mincriterion numeric 0.95 [0,1][0, 1]
logmincriterion numeric -0.05129329 (−∞,∞)(-\infty, \infty)
minsplit integer 20 [1,∞)[1, \infty)
minbucket integer 7 [1,∞)[1, \infty)
minprob numeric 0.01 [0,1][0, 1]
stump logical FALSE TRUE, FALSE -
lookahead logical FALSE TRUE, FALSE -
MIA logical FALSE TRUE, FALSE -
nresample integer 9999 [1,∞)[1, \infty)
tol numeric 1.490116e-08 [0,∞)[0, \infty)
maxsurrogate integer 0 [0,∞)[0, \infty)
numsurrogate logical FALSE TRUE, FALSE -
maxdepth integer Inf [0,∞)[0, \infty)
multiway logical FALSE TRUE, FALSE -
splittry integer 2 [0,∞)[0, \infty)
intersplit logical FALSE TRUE, FALSE -
majority logical FALSE TRUE, FALSE -
caseweights logical TRUE TRUE, FALSE -
saveinfo logical FALSE TRUE, FALSE -
update logical FALSE TRUE, FALSE -
splitflavour character ctree ctree, exhaustive -
maxvar integer - [1,∞)[1, \infty)
OOB logical FALSE TRUE, FALSE -
simplify logical TRUE TRUE, FALSE -
scale logical TRUE TRUE, FALSE -
maxpts integer 25000 (−∞,∞)(-\infty, \infty)
abseps numeric 0.001 [0,∞)[0, \infty)
releps numeric 0 [0,∞)[0, \infty)

Initial parameter values

  • cores: This parameter is initialized to 1 (default is NULL) to avoid threading conflicts with future.

Custom mlr3 parameters

  • mtry:

    • This hyperparameter can alternatively be set via the added hyperparameter mtryratio as mtry = max(ceiling(mtryratio * n_features), 1). Note that mtry and mtryratio are mutually exclusive.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvCForest

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvCForest$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvCForest$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Hothorn T, Zeileis A (2015). “partykit: A Modular Toolkit for Recursive Partytioning in R.” Journal of Machine Learning Research, 16(118), 3905-3909. http://jmlr.org/papers/v16/hothorn15a.html.

Hothorn T, Hornik K, Zeileis A (2006). “Unbiased Recursive Partitioning: A Conditional Inference Framework.” Journal of Computational and Graphical Statistics, 15(3), 651–674. doi:10.1198/106186006x133933, https://doi.org/10.1198/106186006x133933.

See Also

Examples

# Define the Learner
learner = lrn("surv.cforest", ntree = 50L)
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Survival Cox Model with Likelihood Based Boosting Learner

Description

Fit a Survival Cox model with a likelihood based boosting algorithm. Calls CoxBoost::CoxBoost() from package CoxBoost.

Details

Use LearnerSurvCoxboost and LearnerSurvCVCoxboost for Cox boosting without and with internal cross-validation of boosting step number, respectively. Tuning using the internal optimizer in LearnerSurvCVCoxboost may be more efficient when tuning stepno only. However, for tuning multiple hyperparameters, mlr3tuning and LearnerSurvCoxboost will likely give better results.

Prediction types

This learner returns three prediction types, using the internal predict.CoxBoost() function:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation.

  2. crank: same as lp.

  3. distr: a 2d survival matrix, with observations as rows and time points as columns. The internal transformation uses the Breslow estimator to compute the baseline hazard and compose the survival distributions from the lp predictions.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.coxboost")

Meta Information

Parameters

Id Type Default Levels Range
unpen.index untyped - -
standardize logical TRUE TRUE, FALSE -
stepno integer 100 [0,∞)[0, \infty)
penalty numeric - (−∞,∞)(-\infty, \infty)
criterion character pscore pscore, score, hpscore, hscore -
stepsize.factor numeric 1 (−∞,∞)(-\infty, \infty)
sf.scheme character sigmoid sigmoid, linear -
pendistmat untyped - -
connected.index untyped - -
x.is.01 logical FALSE TRUE, FALSE -
return.score logical TRUE TRUE, FALSE -
trace logical FALSE TRUE, FALSE -
at.step untyped - -

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvCoxboost

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvCoxboost$new()

Method selected_features()

Returns the set of selected features which have non-zero coefficients. Calls the internal coef.CoxBoost() function.

Usage
LearnerSurvCoxboost$selected_features(at_step = NULL)
Arguments
at_step

(integer(1))
Which boosting step to get the coefficients for. If no step is given (default), the final boosting step is used.

Returns

(character()) vector of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvCoxboost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Binder, Harald, Schumacher, Martin (2008). “Allowing for mandatory covariates in boosting estimation of sparse high-dimensional survival models.” BMC Bioinformatics, 9(1), 1-10. ISSN 14712105, doi:10.1186/1471-2105-9-14.

See Also

Examples

# Define the Learner
learner = lrn("surv.coxboost")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Survival Conditional Inference Tree Learner

Description

Survival Partition Tree where a significance test is used to determine the univariate splits. Calls partykit::ctree() from partykit.

Prediction types

This learner returns two prediction types:

  1. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using the internal partykit::predict.party() function.

  2. crank: the expected mortality using mlr3proba::surv_return().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.ctree")

Meta Information

Parameters

Id Type Default Levels Range
teststat character quadratic quadratic, maximum -
splitstat character quadratic quadratic, maximum -
splittest logical FALSE TRUE, FALSE -
testtype character Bonferroni Bonferroni, MonteCarlo, Univariate, Teststatistic -
nmax untyped - -
alpha numeric 0.05 [0,1][0, 1]
mincriterion numeric 0.95 [0,1][0, 1]
logmincriterion numeric - (−∞,∞)(-\infty, \infty)
minsplit integer 20 [1,∞)[1, \infty)
minbucket integer 7 [1,∞)[1, \infty)
minprob numeric 0.01 [0,∞)[0, \infty)
stump logical FALSE TRUE, FALSE -
lookahead logical FALSE TRUE, FALSE -
MIA logical FALSE TRUE, FALSE -
nresample integer 9999 [1,∞)[1, \infty)
tol numeric - [0,∞)[0, \infty)
maxsurrogate integer 0 [0,∞)[0, \infty)
numsurrogate logical FALSE TRUE, FALSE -
mtry integer Inf [0,∞)[0, \infty)
maxdepth integer Inf [0,∞)[0, \infty)
maxvar integer - [1,∞)[1, \infty)
multiway logical FALSE TRUE, FALSE -
splittry integer 2 [0,∞)[0, \infty)
intersplit logical FALSE TRUE, FALSE -
majority logical FALSE TRUE, FALSE -
caseweights logical FALSE TRUE, FALSE -
applyfun untyped - -
cores integer NULL (−∞,∞)(-\infty, \infty)
saveinfo logical TRUE TRUE, FALSE -
update logical FALSE TRUE, FALSE -
splitflavour character ctree ctree, exhaustive -
cluster untyped - -
scores untyped - -
doFit logical TRUE TRUE, FALSE -
maxpts integer 25000 (−∞,∞)(-\infty, \infty)
abseps numeric 0.001 [0,∞)[0, \infty)
releps numeric 0 [0,∞)[0, \infty)

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvCTree

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvCTree$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvCTree$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

adibender

References

Hothorn T, Zeileis A (2015). “partykit: A Modular Toolkit for Recursive Partytioning in R.” Journal of Machine Learning Research, 16(118), 3905-3909. http://jmlr.org/papers/v16/hothorn15a.html.

Hothorn T, Hornik K, Zeileis A (2006). “Unbiased Recursive Partitioning: A Conditional Inference Framework.” Journal of Computational and Graphical Statistics, 15(3), 651–674. doi:10.1198/106186006x133933, https://doi.org/10.1198/106186006x133933.

See Also

Examples

# Define the Learner
learner = lrn("surv.ctree")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Survival Cox Model with Cross-Validation Likelihood Based Boosting Learner

Description

Fits a survival Cox model using likelihood based boosting and internal cross-validation for the number of steps. Calls CoxBoost::CoxBoost() or CoxBoost::cv.CoxBoost() from package CoxBoost.

Details

Use LearnerSurvCoxboost and LearnerSurvCVCoxboost for Cox boosting without and with internal cross-validation of boosting step number, respectively. Tuning using the internal optimizer in LearnerSurvCVCoxboost may be more efficient when tuning stepno only. However, for tuning multiple hyperparameters, mlr3tuning and LearnerSurvCoxboost will likely give better results.

If penalty == "optimCoxBoostPenalty" then CoxBoost::optimCoxBoostPenalty() is used to determine the penalty value and optimal step to be used in CoxBoost::CoxBoost().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.cv_coxboost")

Meta Information

Parameters

Id Type Default Levels Range
maxstepno integer 100 [0,∞)[0, \infty)
K integer 10 [2,∞)[2, \infty)
type character verweij verweij, naive -
folds untyped NULL -
minstepno integer 50 [0,∞)[0, \infty)
start.penalty numeric - (−∞,∞)(-\infty, \infty)
iter.max integer 10 [1,∞)[1, \infty)
upper.margin numeric 0.05 [0,1][0, 1]
unpen.index untyped - -
standardize logical TRUE TRUE, FALSE -
penalty numeric - (−∞,∞)(-\infty, \infty)
criterion character pscore pscore, score, hpscore, hscore -
stepsize.factor numeric 1 (−∞,∞)(-\infty, \infty)
sf.scheme character sigmoid sigmoid, linear -
pendistmat untyped - -
connected.index untyped - -
x.is.01 logical FALSE TRUE, FALSE -
return.score logical TRUE TRUE, FALSE -
trace logical FALSE TRUE, FALSE -
at.step untyped - -

Prediction types

This learner returns three prediction types, using the internal predict.CoxBoost() function:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation.

  2. crank: same as lp.

  3. distr: a 2d survival matrix, with observations as rows and time points as columns. The internal transformation uses the Breslow estimator to compute the baseline hazard and compose the survival distributions from the lp predictions.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvCVCoxboost

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvCVCoxboost$new()

Method selected_features()

Returns the set of selected features which have non-zero coefficients. Calls the internal coef.CoxBoost() function.

Usage
LearnerSurvCVCoxboost$selected_features(at_step = NULL)
Arguments
at_step

(integer(1))
Which boosting step to get the coefficients for. If no step is given (default), the final boosting step is used.

Returns

(character()) vector of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvCVCoxboost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Binder, Harald, Schumacher, Martin (2008). “Allowing for mandatory covariates in boosting estimation of sparse high-dimensional survival models.” BMC Bioinformatics, 9(1), 1-10. ISSN 14712105, doi:10.1186/1471-2105-9-14.

See Also

Examples

task = tsk("rats")
task$col_roles$feature = c("litter", "rx")
learner = lrn("surv.cv_coxboost", maxstepno = 20)
splits = partition(task)
learner$train(task, splits$train)
pred = learner$predict(task, splits$test)

Cross-Validated GLM with Elastic Net Regularization Survival Learner

Description

Generalized linear models with elastic net regularization. Calls glmnet::cv.glmnet() from package glmnet.

Initial parameter values

  • family is set to "cox" and cannot be changed.

Prediction types

This learner returns three prediction types:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using glmnet::predict.cv.glmnet().

  2. crank: same as lp.

  3. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using glmnet::survfit.cv.glmnet(). Parameters stype and ctype relate to how lp predictions are transformed into survival predictions and are described in survival::survfit.coxph(). By default the Breslow estimator is used for computing the baseline hazard.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.cv_glmnet")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, glmnet

Parameters

Id Type Default Levels Range
alignment character lambda lambda, fraction -
alpha numeric 1 [0,1][0, 1]
big numeric 9.9e+35 (−∞,∞)(-\infty, \infty)
devmax numeric 0.999 [0,1][0, 1]
dfmax integer - [0,∞)[0, \infty)
eps numeric 1e-06 [0,1][0, 1]
epsnr numeric 1e-08 [0,1][0, 1]
exclude untyped - -
exmx numeric 250 (−∞,∞)(-\infty, \infty)
fdev numeric 1e-05 [0,1][0, 1]
foldid untyped NULL -
gamma untyped - -
grouped logical TRUE TRUE, FALSE -
intercept logical TRUE TRUE, FALSE -
keep logical FALSE TRUE, FALSE -
lambda untyped - -
lambda.min.ratio numeric - [0,1][0, 1]
lower.limits untyped -Inf -
maxit integer 100000 [1,∞)[1, \infty)
mnlam integer 5 [1,∞)[1, \infty)
mxit integer 100 [1,∞)[1, \infty)
mxitnr integer 25 [1,∞)[1, \infty)
nfolds integer 10 [3,∞)[3, \infty)
nlambda integer 100 [1,∞)[1, \infty)
use_pred_offset logical TRUE TRUE, FALSE -
parallel logical FALSE TRUE, FALSE -
penalty.factor untyped - -
pmax integer - [0,∞)[0, \infty)
pmin numeric 1e-09 [0,1][0, 1]
prec numeric 1e-10 (−∞,∞)(-\infty, \infty)
predict.gamma numeric gamma.1se (−∞,∞)(-\infty, \infty)
relax logical FALSE TRUE, FALSE -
s numeric lambda.1se [0,∞)[0, \infty)
standardize logical TRUE TRUE, FALSE -
standardize.response logical FALSE TRUE, FALSE -
thresh numeric 1e-07 [0,∞)[0, \infty)
trace.it integer 0 [0,1][0, 1]
type.gaussian character - covariance, naive -
type.logistic character Newton Newton, modified.Newton -
type.measure character deviance deviance, C -
type.multinomial character ungrouped ungrouped, grouped -
upper.limits untyped Inf -
stype integer 2 [1,2][1, 2]
ctype integer - [1,2][1, 2]

Offset

If a Task contains a column with the offset role, it is automatically incorporated during training via the offset argument in glmnet::glmnet(). During prediction, the offset column from the test set is used only if use_pred_offset = TRUE (default), passed via the newoffset argument in glmnet::predict.coxnet(). Otherwise, if the user sets use_pred_offset = FALSE, a zero offset is applied, effectively disabling the offset adjustment during prediction.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvCVGlmnet

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvCVGlmnet$new()

Method selected_features()

Returns the set of selected features as reported by glmnet::predict.glmnet() with type set to "nonzero".

Usage
LearnerSurvCVGlmnet$selected_features(lambda = NULL)
Arguments
lambda

(numeric(1))
Custom lambda, defaults to the active lambda depending on parameter set.

Returns

(character()) of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvCVGlmnet$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Friedman J, Hastie T, Tibshirani R (2010). “Regularization Paths for Generalized Linear Models via Coordinate Descent.” Journal of Statistical Software, 33(1), 1–22. doi:10.18637/jss.v033.i01.

See Also

Examples

# Define the Learner
learner = lrn("surv.cv_glmnet")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Cross-Validated MCP- or SCAD-penalized Survival Learner

Description

MCP- or SCAD-penalized regression survival model with k-fold cross validation for choosing the regularization parameter lambda. Calls ncvreg::cv.ncvsurv() from package ncvreg. Covariates are internally standardized by ncvreg prior to model fitting.

Prediction types

This learner returns three prediction types (by default the lambda corresponding to the minimum cross-validation error is used):

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using ncvreg::predict.cv.ncvsurv() with type = "link".

  2. crank: identical to lp.

  3. distr: a survival probability matrix with dimensions ⁠n_obs × n_times⁠, where rows correspond to observations and columns correspond to time points. Calculated using ncvreg::predict.cv.ncvsurv() with type = "survival". By default the Kalbfleisch-Prentice estimator is used for the survival function and the time grid is defined by the unique time points in the training data.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.cv_ncvsurv")

Meta Information

Parameters

Id Type Default Levels Range
penalty character MCP MCP, SCAD, lasso -
gamma numeric - [0,∞)[0, \infty)
alpha numeric 1 [0,1][0, 1]
lambda.min numeric - [0,∞)[0, \infty)
nlambda integer 100 [1,∞)[1, \infty)
lambda untyped - -
eps numeric 1e-04 (−∞,∞)(-\infty, \infty)
max.iter integer 10000 [1,∞)[1, \infty)
convex logical TRUE TRUE, FALSE -
dfmax integer - [0,∞)[0, \infty)
penalty.factor untyped - -
warn logical TRUE TRUE, FALSE -
returnX logical TRUE TRUE, FALSE -
cluster untyped - -
nfolds integer 10 (−∞,∞)(-\infty, \infty)
fold untyped - -
se character quick quick, bootstrap -
returnY logical FALSE TRUE, FALSE -
trace logical FALSE TRUE, FALSE -
pred_lambda numeric - (−∞,∞)(-\infty, \infty)

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvNCVsurv

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvNCVsurv$new()

Method selected_features()

Returns the set of selected features which have non-zero coefficients. Calls the internal coef.cv.ncvreg() function.

Usage
LearnerSurvNCVsurv$selected_features(lambda = NULL)
Arguments
lambda

(numeric(1))
Custom lambda, defaults to the lambda with the minimum CV error.

Returns

(character()) vector of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvNCVsurv$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Breheny P, Huang J (2011). “Coordinate descent algorithms for nonconvex penalized regression, with applications to biological feature selection.” Annals of Applied Statistics, 5(1), 232–253. ISSN 1932-6157, doi:10.1214/10-AOAS388.

See Also

Examples

# Define the Learner
learner = lrn("surv.cv_ncvsurv")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Survival Flexible Parametric Learner

Description

Flexible parametric learner. Calls flexsurv::flexsurvreg() from flexsurv.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.flexreg")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”, “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, flexsurv, survival

Parameters

Id Type Default Levels Range
formula untyped - -
anc untyped - -
bhazard untyped - -
rtrunc untyped - -
dist character - gengamma, gengamma.orig, genf, genf.orig, weibull, weibullph, gamma, exp, exponential, llogis, [...] -
inits untyped - -
fixedpars untyped - -
cl numeric 0.95 [0,1][0, 1]
hessian logical TRUE TRUE, FALSE -
hess.control untyped - -
maxiter integer 30 (−∞,∞)(-\infty, \infty)
rel.tolerance numeric 1e-09 (−∞,∞)(-\infty, \infty)
toler.chol numeric 1e-10 (−∞,∞)(-\infty, \infty)
debug integer 0 [0,1][0, 1]
outer.max integer 10 (−∞,∞)(-\infty, \infty)
times untyped - -

Prediction types

This learner returns three prediction types:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using predict.flexsurvreg(type = "lp", ...). The interpretation whether higher lp corresponds to higher risk or higher survival, depends on the distribution choice (see dist parameter documentation).

  2. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using predict.flexsurvreg(type = "survival", ...).

  3. crank: same as lp.

  4. response: mean survival time calculated using predict.flexsurvreg(type = "response", ...).

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvFlexReg

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvFlexReg$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvFlexReg$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Jackson, H. C (2016). “flexsurv: A Platform for Parametric Survival Modeling in R.” Journal of Statistical Software, 70(8), 1–33. doi:10.18637/JSS.V070.I08, https://www.jstatsoft.org/index.php/jss/article/view/v070i08.

See Also

Examples

library(survival)

# Define the task, split to train/test set
task = tsk("lung")
set.seed(42)
part = partition(task)

# Define the learner
learner = lrn("surv.flexreg",
  formula = survival::Surv(time, status) ~ age + ph.karno + sex,
  anc = list(shape = ~ sex), dist = "weibull")

# Train the learner on the training ids
learner$train(task, part$train)
print(learner$model)

# Make predictions for the test rows
predictions = learner$predict(task, part$test)
print(predictions)

# Score the predictions
predictions$score()

Survival Flexible Parametric Spline Learner

Description

Flexible survival regression using the Royston/Parmar spline model. Calls flexsurv::flexsurvspline() from flexsurv.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.flexspline")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”, “response”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, flexsurv, survival

Parameters

Id Type Default Levels Range
formula untyped - -
bhazard untyped - -
rtrunc untyped - -
k integer 0 [0,∞)[0, \infty)
knots untyped - -
bknots untyped - -
scale character hazard hazard, odds, normal -
timescale character log log, identity -
spline character rp rp, splines2ns -
inits untyped - -
fixedpars untyped - -
cl numeric 0.95 [0,1][0, 1]
anc untyped - -
maxiter integer 30 (−∞,∞)(-\infty, \infty)
rel.tolerance numeric 1e-09 (−∞,∞)(-\infty, \infty)
toler.chol numeric 1e-10 (−∞,∞)(-\infty, \infty)
debug integer 0 [0,1][0, 1]
outer.max integer 10 (−∞,∞)(-\infty, \infty)
times untyped - -

Prediction types

This learner returns three prediction types:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using predict.flexsurvreg(type = "lp", ...). Higher linear predictor gives higher hazard (equivalently, lower survival) for the spline models.

  2. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using predict.flexsurvreg(type = "survival", ...).

  3. crank: same as lp.

  4. response: mean survival time calculated using predict.flexsurvreg(type = "response", ...)

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvFlexSpline

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvFlexSpline$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvFlexSpline$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Royston, Patrick, Parmar, KB M (2002). “Flexible parametric proportional-hazards and proportional-odds models for censored survival data, with application to prognostic modelling and estimation of treatment effects.” Statistics in medicine, 21(15), 2175–2197.

See Also

Examples

library(survival)

# Define the task, split to train/test set
task = tsk("lung")
set.seed(42)
part = partition(task)

# Define the learner
learner = lrn("surv.flexspline", k = 1,
  formula = survival::Surv(time, status) ~ age + ph.karno + sex,
  anc = list(gamma1 = ~ sex))

# Train the learner on the training ids
learner$train(task, part$train)
print(learner$model)

# Make predictions for the test rows
predictions = learner$predict(task, part$test)
print(predictions)

# Score the predictions
predictions$score()

Cox Proportional Hazards Generalized Additive Learner

Description

Cox proportional hazards model fitted via generalized additive modeling. Internally calls mgcv::gam() from the mgcv package with family = mgcv::cox.ph(). The censoring indicator is passed via the weights argument.

Prediction types

Three types of prediction are returned for this learner:

  1. lp: a vector of linear predictors (relative risk scores), one per observation. Calls mgcv::predict.gam() with type = "link".

  2. crank: same as lp.

  3. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. By default, the Breslow estimator is used via mlr3proba::breslow().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.gam.cox")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, mgcv

Parameters

Id Type Default Levels Range
formula untyped - -
method character GCV.Cp GCV.Cp, GACV.Cp, REML, P-REML, ML, P-ML -
optimizer untyped c("outer", "newton") -
scale numeric 0 (−∞,∞)(-\infty, \infty)
select logical FALSE TRUE, FALSE -
knots untyped NULL -
sp untyped NULL -
min.sp untyped NULL -
H untyped NULL -
gamma numeric 1 [1,∞)[1, \infty)
paraPen untyped NULL -
G untyped NULL -
in.out untyped NULL -
drop.unused.levels logical TRUE TRUE, FALSE -
drop.intercept logical FALSE TRUE, FALSE -
nei untyped - -
nthreads integer 1 [1,∞)[1, \infty)
ncv.threads integer 1 [1,∞)[1, \infty)
irls.reg numeric 0 [0,∞)[0, \infty)
epsilon numeric 1e-07 [0,∞)[0, \infty)
maxit integer 200 (−∞,∞)(-\infty, \infty)
mgcv.tol numeric 1e-07 [0,∞)[0, \infty)
mgcv.half integer 15 [0,∞)[0, \infty)
trace logical FALSE TRUE, FALSE -
rank.tol numeric 1.490116e-08 [0,∞)[0, \infty)
nlm untyped list() -
optim untyped list() -
newton untyped list() -
idLinksBases logical TRUE TRUE, FALSE -
scalePenalty logical TRUE TRUE, FALSE -
efs.lspmax integer 15 [0,∞)[0, \infty)
efs.tol numeric 0.1 [0,∞)[0, \infty)
scale.est character fletcher fletcher, pearson, deviance -
edge.correct logical FALSE TRUE, FALSE -
block.size integer 1000 (−∞,∞)(-\infty, \infty)
unconditional logical FALSE TRUE, FALSE -

Formula

A gam formula specific to the task at hand is required for the formula parameter (see example and ?mgcv::formula.gam). Beware, if no formula is provided, a fallback formula is used that will make the gam behave like a glm (this behavior is required for the unit tests). Only features specified in the formula will be used, superseding columns with col_roles "feature" in the task.

Offset

If a Task contains a column with the offset role, it is automatically incorporated during training via the offset argument in mgcv::gam(). No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvGamCox

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvGamCox$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvGamCox$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

bblodfon

References

Wood, Simon (2012). “mgcv: Mixed GAM Computation Vehicle with GCV/AIC/REML smoothness estimation.”

Wood, N. S, Pya, Natalya, Safken, Benjamin (2016). “Smoothing Parameter and Model Selection for General Smooth Models.” Journal of the American Statistical Association, 111(516), 1548–1563. doi:10.1080/01621459.2016.1180986.

Examples

# simple example
t = tsk("lung")
l = lrn("surv.gam.cox")
l$param_set$set_values(formula = time ~ s(age, k = 5) + ph.karno + sex)
l$train(t)
l$model

Boosted Generalized Additive Survival Learner

Description

Fits a generalized additive survival model using a boosting algorithm. Calls mboost::gamboost() from mboost.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.gamboost")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, mboost

Parameters

Id Type Default Levels Range
family character coxph coxph, weibull, loglog, lognormal, gehan, cindex, custom -
custom.family untyped - -
nuirange untyped c(0, 100) -
center logical TRUE TRUE, FALSE -
mstop integer 100 [0,∞)[0, \infty)
nu numeric 0.1 [0,1][0, 1]
risk character inbag inbag, oobag, none -
stopintern untyped FALSE -
trace logical FALSE TRUE, FALSE -
oobweights untyped NULL -
baselearner character bbs bbs, bols, btree -
dfbase integer 4 [0,∞)[0, \infty)
sigma numeric 0.1 [0,1][0, 1]
ipcw untyped 1 -
na.action untyped stats::na.omit -

Prediction types

This learner returns two to three prediction types:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using mboost::predict.gamboost(). If the family parameter is not "coxph", -lp is returned, since non-coxph families represent AFT-style distributions where lower lp values indicate higher risk.

  2. crank: same as lp.

  3. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using mboost::survFit(). This prediction type is present only when the family distribution parameter is equal to "coxph" (default). By default the Breslow estimator is used for computing the baseline hazard.

Offset

If a Task contains a column with the offset role, it is automatically incorporated via the offset argument in mboost's training function. No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvGAMBoost

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvGAMBoost$new()

Method importance()

The importance scores are extracted with the function mboost::varimp() with the default arguments.

Usage
LearnerSurvGAMBoost$importance()
Returns

Named numeric().

Method selected_features()

Selected features are extracted with the function mboost::variable.names.mboost(), with used.only = TRUE.

Usage
LearnerSurvGAMBoost$selected_features()
Returns

character().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvGAMBoost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Bühlmann, Peter, Yu, Bin (2003). “Boosting with the L 2 loss: regression and classification.” Journal of the American Statistical Association, 98(462), 324–339.

See Also

Examples

lrn("surv.gamboost")

Survival Gradient Boosting Machine Learner

Description

Gradient Boosting for Survival Analysis. Calls gbm::gbm() from gbm.

Prediction types

This learner returns two prediction types, using the internal predict.gbm() function:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation.

  2. crank: same as lp.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.gbm")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “lp”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, gbm

Parameters

Id Type Default Levels Range
distribution character coxph coxph -
n.trees integer 100 [1,∞)[1, \infty)
cv.folds integer 0 [0,∞)[0, \infty)
interaction.depth integer 1 [1,∞)[1, \infty)
n.minobsinnode integer 10 [1,∞)[1, \infty)
shrinkage numeric 0.001 [0,∞)[0, \infty)
bag.fraction numeric 0.5 [0,1][0, 1]
train.fraction numeric 1 [0,1][0, 1]
keep.data logical FALSE TRUE, FALSE -
verbose logical FALSE TRUE, FALSE -
var.monotone untyped - -
n.cores integer 1 (−∞,∞)(-\infty, \infty)
single.tree logical FALSE TRUE, FALSE -

Initial parameter values

  • distribution:

  • Actual default: "bernoulli"

  • Adjusted default: "coxph"

  • Reason for change: This is the only distribution available for survival.

  • keep.data:

    • Actual default: TRUE

    • Adjusted default: FALSE

    • Reason for change: keep.data = FALSE saves memory during model fitting.

  • n.cores:

    • Actual default: NULL

    • Adjusted default: 1

    • Reason for change: Suppressing the automatic internal parallelization if cv.folds > 0 and avoid threading conflicts with future.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvGBM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvGBM$new()

Method importance()

The importance scores are extracted from the model slot variable.importance.

Usage
LearnerSurvGBM$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvGBM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Friedman, H J (2002). “Stochastic gradient boosting.” Computational statistics & data analysis, 38(4), 367–378.

See Also

Examples

# Define the Learner
learner = lrn("surv.gbm")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Boosted Generalized Linear Survival Learner

Description

Fits a generalized linear survival model using a boosting algorithm. Calls mboost::glmboost() from mboost.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.glmboost")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, mboost

Parameters

Id Type Default Levels Range
family character coxph coxph, weibull, loglog, lognormal, gehan, cindex, custom -
custom.family untyped - -
nuirange untyped c(0, 100) -
center logical TRUE TRUE, FALSE -
mstop integer 100 [0,∞)[0, \infty)
nu numeric 0.1 [0,1][0, 1]
risk character inbag inbag, oobag, none -
oobweights untyped NULL -
stopintern logical FALSE TRUE, FALSE -
trace logical FALSE TRUE, FALSE -
sigma numeric 0.1 [0,1][0, 1]
ipcw untyped 1 -
na.action untyped stats::na.omit -
contrasts.arg untyped - -

Prediction types

This learner returns two to three prediction types:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using mboost::predict.glmboost(). If the family parameter is not "coxph", -lp is returned, since non-coxph families represent AFT-style distributions where lower lp values indicate higher risk.

  2. crank: same as lp.

  3. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using mboost::survFit(). This prediction type is present only when the family distribution parameter is equal to "coxph" (default). By default the Breslow estimator is used for computing the baseline hazard.

Offset

If a Task contains a column with the offset role, it is automatically incorporated via the offset argument in mboost's training function. No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvGLMBoost

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvGLMBoost$new()

Method importance()

Importance scores are extracted with the function mboost::varimp() and represent a feature's individual contribution to the risk reduction per boosting step of the fitted model. The higher the risk reduction, the larger the feature importance.

Note: Importance is supported only for datasets with numeric features, as the presence of factors with multiple levels makes it difficult to get the original feature names.

Usage
LearnerSurvGLMBoost$importance()
Returns

Named numeric().

Method selected_features()

Selected features are extracted with the function mboost::coef.glmboost() which by default returns features with non-zero coefficients and for the final number of boosting iterations.

Note: Selected features can be retrieved only for datasets with numeric features, as the presence of factors with multiple levels makes it difficult to get the original feature names.

Usage
LearnerSurvGLMBoost$selected_features()
Returns

character().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvGLMBoost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Bühlmann, Peter, Yu, Bin (2003). “Boosting with the L 2 loss: regression and classification.” Journal of the American Statistical Association, 98(462), 324–339.

See Also

Examples

# Define the Learner
learner = lrn("surv.glmboost")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

GLM with Elastic Net Regularization Survival Learner

Description

Generalized linear models with elastic net regularization. Calls glmnet::glmnet() from package glmnet.

Initial parameter values

  • family is set to "cox" and cannot be changed.

Prediction types

This learner returns three prediction types:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using glmnet::predict.coxnet().

  2. crank: same as lp.

  3. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using glmnet::survfit.coxnet(). Parameters stype and ctype relate to how lp predictions are transformed into survival predictions and are described in survival::survfit.coxph(). By default the Breslow estimator is used for computing the baseline hazard.

Caution: This learner is different to learners calling glmnet::cv.glmnet() in that it does not use the internal optimization of parameter lambda. Instead, lambda needs to be tuned by the user (e.g., via mlr3tuning). When lambda is tuned, the glmnet will be trained for each tuning iteration. While fitting the whole path of lambdas would be more efficient, as is done by default in glmnet::glmnet(), tuning/selecting the parameter at prediction time (using parameter s) is currently not supported in mlr3 (at least not in efficient manner). Tuning the s parameter is, therefore, currently discouraged.

When the data are i.i.d. and efficiency is key, we recommend using the respective auto-tuning counterpart in mlr_learners_surv.cv_glmnet(). However, in some situations this is not applicable, usually when data are imbalanced or not i.i.d. (longitudinal, time-series) and tuning requires custom resampling strategies (blocked design, stratification).

Offset

If a Task contains a column with the offset role, it is automatically incorporated during training via the offset argument in glmnet::glmnet(). During prediction, the offset column from the test set is used only if use_pred_offset = TRUE (default), passed via the newoffset argument in glmnet::predict.coxnet(). Otherwise, if the user sets use_pred_offset = FALSE, a zero offset is applied, effectively disabling the offset adjustment during prediction.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.glmnet")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, glmnet

Parameters

Id Type Default Levels Range
alignment character lambda lambda, fraction -
alpha numeric 1 [0,1][0, 1]
big numeric 9.9e+35 (−∞,∞)(-\infty, \infty)
devmax numeric 0.999 [0,1][0, 1]
dfmax integer - [0,∞)[0, \infty)
eps numeric 1e-06 [0,1][0, 1]
epsnr numeric 1e-08 [0,1][0, 1]
exact logical FALSE TRUE, FALSE -
exclude untyped - -
exmx numeric 250 (−∞,∞)(-\infty, \infty)
fdev numeric 1e-05 [0,1][0, 1]
gamma untyped - -
grouped logical TRUE TRUE, FALSE -
intercept logical TRUE TRUE, FALSE -
keep logical FALSE TRUE, FALSE -
lambda untyped - -
lambda.min.ratio numeric - [0,1][0, 1]
lower.limits untyped -Inf -
maxit integer 100000 [1,∞)[1, \infty)
mnlam integer 5 [1,∞)[1, \infty)
mxit integer 100 [1,∞)[1, \infty)
mxitnr integer 25 [1,∞)[1, \infty)
nlambda integer 100 [1,∞)[1, \infty)
use_pred_offset logical TRUE TRUE, FALSE -
parallel logical FALSE TRUE, FALSE -
penalty.factor untyped - -
pmax integer - [0,∞)[0, \infty)
pmin numeric 1e-09 [0,1][0, 1]
prec numeric 1e-10 (−∞,∞)(-\infty, \infty)
predict.gamma numeric gamma.1se (−∞,∞)(-\infty, \infty)
relax logical FALSE TRUE, FALSE -
s numeric 0.01 [0,∞)[0, \infty)
standardize logical TRUE TRUE, FALSE -
thresh numeric 1e-07 [0,∞)[0, \infty)
trace.it integer 0 [0,1][0, 1]
type.logistic character Newton Newton, modified.Newton -
type.multinomial character ungrouped ungrouped, grouped -
upper.limits untyped Inf -
stype integer 2 [1,2][1, 2]
ctype integer - [1,2][1, 2]

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvGlmnet

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvGlmnet$new()

Method selected_features()

Returns the set of selected features as reported by glmnet::predict.glmnet() with type set to "nonzero".

Usage
LearnerSurvGlmnet$selected_features(lambda = NULL)
Arguments
lambda

(numeric(1))
Custom lambda, defaults to the active lambda depending on parameter set.

Returns

(character()) of feature names.

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvGlmnet$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Friedman J, Hastie T, Tibshirani R (2010). “Regularization Paths for Generalized Linear Models via Coordinate Descent.” Journal of Statistical Software, 33(1), 1–22. doi:10.18637/jss.v033.i01.

See Also

Examples

# Define the Learner
learner = lrn("surv.glmnet")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Boosted Generalized Additive Survival Learner

Description

Model-based boosting for survival analysis. Calls mboost::mboost() from mboost.

Details

distr prediction made by mboost::survFit().

Prediction types

This learner returns two to three prediction types:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using mboost::predict.mboost(). If the family parameter is not "coxph", -lp is returned, since non-coxph families represent AFT-style distributions where lower lp values indicate higher risk.

  2. crank: same as lp.

  3. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using mboost::survFit(). This prediction type is present only when the family distribution parameter is equal to "coxph" (default). By default the Breslow estimator is used for computing the baseline hazard.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.mboost")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, mboost

Parameters

Id Type Default Levels Range
family character coxph coxph, weibull, loglog, lognormal, gehan, cindex, custom -
custom.family untyped - -
nuirange untyped c(0, 100) -
center logical TRUE TRUE, FALSE -
mstop integer 100 [0,∞)[0, \infty)
nu numeric 0.1 [0,1][0, 1]
risk character inbag inbag, oobag, none -
stopintern logical FALSE TRUE, FALSE -
trace logical FALSE TRUE, FALSE -
oobweights untyped NULL -
baselearner character bbs bbs, bols, btree -
sigma numeric 0.1 [0,1][0, 1]
ipcw untyped 1 -
na.action untyped stats::na.omit -

Offset

If a Task contains a column with the offset role, it is automatically incorporated via the offset argument in mboost's training function. No offset is applied during prediction for this learner.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvMBoost

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvMBoost$new()

Method importance()

The importance scores are extracted with the function mboost::varimp() with the default arguments.

Usage
LearnerSurvMBoost$importance()
Returns

Named numeric().

Method selected_features()

Selected features are extracted with the function mboost::variable.names.mboost(), with used.only = TRUE.

Usage
LearnerSurvMBoost$selected_features()
Returns

character().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvMBoost$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Bühlmann, Peter, Yu, Bin (2003). “Boosting with the L 2 loss: regression and classification.” Journal of the American Statistical Association, 98(462), 324–339.

See Also

Examples

lrn("surv.mboost")

Survival Nelson-Aalen Estimator Learner

Description

Non-parametric estimator of the cumulative hazard rate function. Calls survival::survfit() from survival.

Prediction types

This learner returns two prediction types:

  1. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. The cumulative hazard H(t)H(t) is calculated using the train data and the default parameters of the survival::survfit() function, i.e. ctype = 1, which uses the Nelson-Aalen formula. Then for each test observation the survival curve is S(t)=exp⁡(−H(t))S(t) = \exp{(-H(t))}.

  2. crank: the expected mortality using mlr3proba::surv_return().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.nelson")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, survival

Parameters

Empty ParamSet

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvNelson

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvNelson$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvNelson$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Nelson, Wayne (1969). “Hazard plotting for incomplete failure data.” Journal of Quality Technology, 1(1), 27–52.

Nelson, Wayne (1972). “Theory and applications of hazard plotting for censored failure data.” Technometrics, 14(4), 945–966.

Aalen, Odd (1978). “Nonparametric inference for a family of counting processes.” The Annals of Statistics, 701–726.

See Also

Examples

# Define the Learner
learner = lrn("surv.nelson")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Survival L1 and L2 Penalized Cox Learner

Description

Penalized (L1 and L2) Cox Proportional Hazards model. Calls penalized::penalized() from penalized.

Details

The penalized and unpenalized arguments in the learner are implemented slightly differently than in penalized::penalized(). Here, there is no parameter for penalized but instead it is assumed that every variable is penalized unless stated in the unpenalized parameter.

Prediction types

This learner returns two prediction types:

  1. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using the internal penalized::predict() function. By default the Breslow estimator penalized::breslow() is used for computing the baseline hazard.

  2. crank: the expected mortality using mlr3proba::surv_return().

Initial parameter values

  • trace is set to "FALSE" to disable printing output during model training.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.penalized")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”

  • Feature Types: “logical”, “integer”, “numeric”, “factor”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, penalized

Parameters

Id Type Default Levels Range
unpenalized untyped - -
lambda1 untyped 0 -
lambda2 untyped 0 -
positive logical FALSE TRUE, FALSE -
fusedl logical FALSE TRUE, FALSE -
startbeta numeric - (−∞,∞)(-\infty, \infty)
startgamma numeric - (−∞,∞)(-\infty, \infty)
steps integer 1 [1,∞)[1, \infty)
epsilon numeric 1e-10 [0,1][0, 1]
maxiter integer - [1,∞)[1, \infty)
standardize logical FALSE TRUE, FALSE -
trace logical TRUE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvPenalized

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvPenalized$new()

Method selected_features()

Selected features are extracted with the method coef() of the S4 model object, see penalized::penfit(). By default it returns features with non-zero coefficients.

Note: Selected features can be retrieved only for datasets with numeric features, as the presence of factors with multiple levels makes it difficult to get the original feature names.

Usage
LearnerSurvPenalized$selected_features()
Returns

character().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvPenalized$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Goeman, J J (2010). “L1 penalized estimation in the Cox proportional hazards model.” Biometrical journal, 52(1), 70–84.

See Also

Examples

# Define the Learner
learner = lrn("surv.penalized")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)


# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Survival Priority Lasso Learner

Description

Patient outcome prediction based on multi-omics data taking practitioners’ preferences into account. Calls prioritylasso::prioritylasso() from prioritylasso. Many parameters for this survival learner are the same as mlr_learners_surv.cv_glmnet as prioritylasso() calls glmnet::cv.glmnet() during training phase. Note that prioritylasso() has ways to deal with block-wise missing data, but this feature is not supported currently.

Prediction types

This learner returns three prediction types:

  1. lp: a vector containing the linear predictors (relative risk scores), where each score corresponds to a specific test observation. Calculated using prioritylasso::predict.prioritylasso().

  2. crank: same as lp.

  3. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. Calculated using mlr3proba::breslow() where the Breslow estimator is used for computing the baseline hazard.

Initial parameter values

  • family is set to "cox" for the Cox survival objective and cannot be changed

  • type.measure set to "deviance" (cross-validation measure)

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.priority_lasso")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”

  • Feature Types: “logical”, “integer”, “numeric”

  • Required Packages: mlr3, mlr3proba, prioritylasso

Parameters

Id Type Default Levels Range
blocks untyped - -
max.coef untyped NULL -
block1.penalization logical TRUE TRUE, FALSE -
lambda.type character lambda.min lambda.min, lambda.1se -
standardize logical TRUE TRUE, FALSE -
nfolds integer 5 [1,∞)[1, \infty)
foldid untyped NULL -
cvoffset logical FALSE TRUE, FALSE -
cvoffsetnfolds integer 10 [1,∞)[1, \infty)
return.x logical TRUE TRUE, FALSE -
include.allintercepts logical FALSE TRUE, FALSE -
use.blocks untyped "all" -
alignment character lambda lambda, fraction -
alpha numeric 1 [0,1][0, 1]
big numeric 9.9e+35 (−∞,∞)(-\infty, \infty)
devmax numeric 0.999 [0,1][0, 1]
dfmax integer - [0,∞)[0, \infty)
eps numeric 1e-06 [0,1][0, 1]
epsnr numeric 1e-08 [0,1][0, 1]
exclude untyped - -
exmx numeric 250 (−∞,∞)(-\infty, \infty)
fdev numeric 1e-05 [0,1][0, 1]
gamma untyped - -
grouped logical TRUE TRUE, FALSE -
intercept logical TRUE TRUE, FALSE -
keep logical FALSE TRUE, FALSE -
lambda untyped - -
lambda.min.ratio numeric - [0,1][0, 1]
lower.limits untyped -Inf -
maxit integer 100000 [1,∞)[1, \infty)
mnlam integer 5 [1,∞)[1, \infty)
mxit integer 100 [1,∞)[1, \infty)
mxitnr integer 25 [1,∞)[1, \infty)
nlambda integer 100 [1,∞)[1, \infty)
offset untyped NULL -
parallel logical FALSE TRUE, FALSE -
penalty.factor untyped - -
pmax integer - [0,∞)[0, \infty)
pmin numeric 1e-09 [0,1][0, 1]
prec numeric 1e-10 (−∞,∞)(-\infty, \infty)
standardize.response logical FALSE TRUE, FALSE -
thresh numeric 1e-07 [0,∞)[0, \infty)
trace.it integer 0 [0,1][0, 1]
type.gaussian character - covariance, naive -
type.logistic character Newton Newton, modified.Newton -
type.multinomial character ungrouped ungrouped, grouped -
upper.limits untyped Inf -
relax logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvPriorityLasso

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvPriorityLasso$new()

Method selected_features()

Selected features, i.e. those where the coefficient is non-zero.

Usage
LearnerSurvPriorityLasso$selected_features()
Returns

character().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvPriorityLasso$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

HarutyunyanLiana

References

Simon K, Vindi J, Roman H, Tobias H, Anne-Laure B (2018). “Priority-Lasso: a simple hierarchical approach to the prediction of clinical outcome using multi-omics data.” BMC Bioinformatics, 19. doi:10.1186/s12859-018-2344-6.

See Also

Examples

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# check task's features
task$feature_names

# partition features to 2 blocks
blocks = list(bl1 = 1:3, bl2 = 4:6)

# define learner
learner = lrn("surv.priority_lasso", blocks = blocks, block1.penalization = FALSE,
              lambda.type = "lambda.1se", standardize = TRUE, nfolds = 5)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# selected features
learner$selected_features()

# Make predictions for the test observations
pred = learner$predict(task, row_ids = ids$test)
pred

# Score the predictions
pred$score()

Ranger Survival Learner

Description

Random survival forest. Calls ranger::ranger() from package ranger.

Prediction types

This learner returns two prediction types:

  1. distr: a survival matrix in two dimensions, where observations are represented in rows and (unique event) time points in columns. Calculated using the internal ranger::predict.ranger() function.

  2. crank: the expected mortality using mlr3proba::surv_return().

Custom mlr3 parameters

  • mtry: This hyperparameter can alternatively be set via our hyperparameter mtry.ratio as mtry = max(ceiling(mtry.ratio * n_features), 1). Note that mtry and mtry.ratio are mutually exclusive.

Initial parameter values

  • num.threads is initialized to 1 to avoid conflicts with parallelization via future.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.ranger")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, ranger

Parameters

Id Type Default Levels Range
alpha numeric 0.5 (−∞,∞)(-\infty, \infty)
always.split.variables untyped - -
holdout logical FALSE TRUE, FALSE -
importance character - none, impurity, impurity_corrected, permutation -
keep.inbag logical FALSE TRUE, FALSE -
max.depth integer NULL [0,∞)[0, \infty)
min.node.size integer 5 [1,∞)[1, \infty)
minprop numeric 0.1 (−∞,∞)(-\infty, \infty)
mtry integer - [1,∞)[1, \infty)
mtry.ratio numeric - [0,1][0, 1]
num.random.splits integer 1 [1,∞)[1, \infty)
num.threads integer 1 [1,∞)[1, \infty)
num.trees integer 500 [1,∞)[1, \infty)
oob.error logical TRUE TRUE, FALSE -
regularization.factor untyped 1 -
regularization.usedepth logical FALSE TRUE, FALSE -
replace logical TRUE TRUE, FALSE -
respect.unordered.factors character ignore ignore, order, partition -
sample.fraction numeric - [0,1][0, 1]
save.memory logical FALSE TRUE, FALSE -
scale.permutation.importance logical FALSE TRUE, FALSE -
seed integer NULL (−∞,∞)(-\infty, \infty)
split.select.weights numeric - [0,1][0, 1]
splitrule character logrank logrank, extratrees, C, maxstat -
verbose logical TRUE TRUE, FALSE -
write.forest logical TRUE TRUE, FALSE -
min.bucket integer 3 (−∞,∞)(-\infty, \infty)
time.interest integer NULL [1,∞)[1, \infty)
node.stats logical FALSE TRUE, FALSE -

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvRanger

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvRanger$new()

Method importance()

The importance scores are extracted from the model slot variable.importance.

Usage
LearnerSurvRanger$importance()
Returns

Named numeric().

Method oob_error()

The out-of-bag error is extracted from the model slot prediction.error.

Usage
LearnerSurvRanger$oob_error()
Returns

numeric(1).

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvRanger$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

be-marc

References

Wright, N. M, Ziegler, Andreas (2017). “ranger: A Fast Implementation of Random Forests for High Dimensional Data in C++ and R.” Journal of Statistical Software, 77(1), 1–17. doi:10.18637/jss.v077.i01.

Breiman, Leo (2001). “Random Forests.” Machine Learning, 45(1), 5–32. ISSN 1573-0565, doi:10.1023/A:1010933404324.

See Also

Examples

# Define the Learner
learner = lrn("surv.ranger", importance = "permutation")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Survival Random Forest SRC Learner

Description

Random survival forest. Calls randomForestSRC::rfsrc() from randomForestSRC.

Prediction types

This learner returns two prediction types:

  1. distr: a survival matrix in two dimensions, where observations are represented in rows and (unique event) time points in columns. Calculated using the internal randomForestSRC::predict.rfsrc() function.

  2. crank: the expected mortality using mlr3proba::surv_return().

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.rfsrc")

Meta Information

Parameters

Id Type Default Levels Range
ntree integer 500 [1,∞)[1, \infty)
mtry integer - [1,∞)[1, \infty)
mtry.ratio numeric - [0,1][0, 1]
nodesize integer 15 [1,∞)[1, \infty)
nodedepth integer - [1,∞)[1, \infty)
splitrule character logrank logrank, bs.gradient -
nsplit integer 10 [0,∞)[0, \infty)
importance character FALSE FALSE, TRUE, none, anti, permute, random -
block.size integer 10 [1,∞)[1, \infty)
bootstrap character by.root by.root, by.node, none, by.user -
samptype character swor swor, swr -
samp untyped - -
membership logical FALSE TRUE, FALSE -
sampsize untyped - -
sampsize.ratio numeric - [0,1][0, 1]
na.action character na.omit na.omit, na.impute -
nimpute integer 1 [1,∞)[1, \infty)
ntime integer 150 [0,∞)[0, \infty)
proximity character FALSE FALSE, TRUE, inbag, oob, all -
distance character FALSE FALSE, TRUE, inbag, oob, all -
forest.wt character FALSE FALSE, TRUE, inbag, oob, all -
xvar.wt untyped - -
split.wt untyped - -
forest logical TRUE TRUE, FALSE -
var.used character FALSE FALSE, all.trees -
split.depth character FALSE FALSE, all.trees, by.tree -
seed integer - (−∞,−1](-\infty, -1]
do.trace logical FALSE TRUE, FALSE -
get.tree untyped - -
outcome character train train, test -
ptn.count integer 0 [0,∞)[0, \infty)
estimator character nelson nelson, kaplan -
cores integer 1 [1,∞)[1, \infty)
use.uno logical TRUE TRUE, FALSE -
save.memory logical FALSE TRUE, FALSE -
perf.type character - none -
case.depth logical FALSE TRUE, FALSE -
marginal.xvar untyped NULL -

Custom mlr3 parameters

  • estimator: Hidden parameter that controls the type of estimator used to derive the survival function during prediction. The default value is "chf" which uses a bootstrapped Nelson-Aalen estimator for the cumulative hazard function H(t)H(t), (Ishwaran, 2008) from which we calculate S(t)=exp⁡(−H(t))S(t) = \exp(-H(t)), whereas "surv" uses a bootstrapped Kaplan-Meier estimator to directly estimate S(t)S(t).

  • mtry: This hyperparameter can alternatively be set via the added hyperparameter mtry.ratio as mtry = max(ceiling(mtry.ratio * n_features), 1). Note that mtry and mtry.ratio are mutually exclusive.

  • sampsize: This hyperparameter can alternatively be set via the added hyperparameter sampsize.ratio as sampsize = max(ceiling(sampsize.ratio * n_obs), 1). Note that sampsize and sampsize.ratio are mutually exclusive.

  • cores: This value is set as the option rf.cores during training and is set to 1 by default.

Initial parameter values

  • ntime: Number of time points to coerce the observed event times for use in the estimated survival function during prediction. We changed the default value of 150 to 0 in order to be in line with other random survival forest learners and use all the unique event times from the train set.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvRandomForestSRC

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvRandomForestSRC$new()

Method importance()

The importance scores are extracted from the model slot importance.

Usage
LearnerSurvRandomForestSRC$importance()
Returns

Named numeric().

Method selected_features()

Selected features are extracted from the model slot var.used.

Note: Due to a known issue in randomForestSRC, enabling var.used = "all.trees" causes prediction to fail. Therefore, this setting should be used exclusively for feature selection purposes and not when prediction is required.

Usage
LearnerSurvRandomForestSRC$selected_features()
Returns

character().

Method oob_error()

OOB error extracted from the model slot err.rate.

Usage
LearnerSurvRandomForestSRC$oob_error()
Returns

numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvRandomForestSRC$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Ishwaran H, Kogalur UB, Blackstone EH, Lauer MS (2008). “Random survival forests.” The Annals of Applied Statistics, 2(3). doi:10.1214/08-aoas169, https://doi.org/10.1214/08-aoas169.

Breiman, Leo (2001). “Random Forests.” Machine Learning, 45(1), 5–32. ISSN 1573-0565, doi:10.1023/A:1010933404324.

See Also

Examples

# Define the Learner
learner = lrn("surv.rfsrc", importance = "TRUE")
print(learner)

# Define a Task
task = tsk("grace")
# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

SurvDNN Survival Learner

Description

Deep neural network survival models from package survdnn, aimed at tabular (low to moderate-dimensional) covariate settings using torch-based multilayer perceptrons. The learner wraps survdnn::survdnn().

Prediction types

This learner supports the following prediction types:

lp

A numeric vector of linear predictors, one per observation. For loss "cox" / "cox_l2" this is a log-risk score (higher implies worse prognosis). For "aft", predict.survdnn() returns the predicted log-time location μ(x)\mu(x) (higher implies better prognosis), therefore the learner internally negates it such that higher values imply higher risk (consistent with mlr3 conventions). For "coxtime", this is g(t0,x)g(t_0, x) evaluated at a reference time.

crank

same as lp.

distr

A survival matrix (rows = observations, columns = time points) based on predict(type = "survival"). By default, predictions are evaluated on the unique event times of the training data.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.survdnn")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “distr”, “lp”

  • Feature Types: “integer”, “numeric”, “factor”, “ordered”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, survdnn, torch

Parameters

Id Type Default Levels Range
hidden untyped c(32L, 16L) -
activation character relu relu, leaky_relu, tanh, sigmoid, gelu, elu, softplus -
lr numeric 1e-04 [1e−06,1][1e-06, 1]
epochs integer 300 [1,∞)[1, \infty)
loss character cox cox, cox_l2, aft, coxtime -
optimizer character adam adam, adamw, sgd, rmsprop, adagrad -
optim_args untyped list() -
verbose logical FALSE TRUE, FALSE -
dropout numeric 0.3 [0,1][0, 1]
batch_norm logical TRUE TRUE, FALSE -
callbacks untyped NULL -
.seed integer NULL (−∞,∞)(-\infty, \infty)
.device character auto auto, cpu, cuda -
na_action character omit omit, fail -

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvDNN

Active bindings

marshaled

(logical(1)) Whether the learner has been marshaled.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvDNN$new()

Method marshal()

Marshal the learner's model.

Usage
LearnerSurvDNN$marshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::marshal_model().

Method unmarshal()

Unmarshal the learner's model.

Usage
LearnerSurvDNN$unmarshal(...)
Arguments
...

(any)
Additional arguments passed to mlr3::unmarshal_model().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvDNN$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

ielbadisy

See Also

Examples

# Define the Learner
learner = lrn("surv.survdnn", epochs = 42L, loss = "cox")
print(learner)

# Define the task, split to train/test set
task = tsk("lung")
set.seed(42)
part = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = part$train)
print(learner$model)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = part$test)
print(predictions)

# Score the predictions
predictions$score()

Survival Support Vector Machine Learner

Description

Survival support vector machine. Calls survivalsvm::survivalsvm() from survivalsvm.

Details

Four possible SVMs can be implemented, dependent on the type parameter. These correspond to predicting the survival time via regression (regression), predicting a continuous rank (vanbelle1, vanbelle2), or a hybrid of the two (hybrid). Whichever type is chosen determines how the crank predict type is calculated, but in any case all can be considered a valid continuous ranking.

makediff3 is recommended when using type = "hybrid".

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.svm")

Meta Information

  • Task type: “surv”

  • Predict Types: “crank”, “response”

  • Feature Types: “logical”, “integer”, “numeric”, “character”, “factor”

  • Required Packages: mlr3, mlr3proba, mlr3extralearners, survivalsvm

Parameters

Id Type Default Levels Range
type character regression regression, vanbelle1, vanbelle2, hybrid -
diff.meth character - makediff1, makediff2, makediff3 -
gamma numeric NULL (−∞,∞)(-\infty, \infty)
mu numeric NULL (−∞,∞)(-\infty, \infty)
opt.meth character quadprog quadprog, ipop -
kernel character lin_kernel lin_kernel, add_kernel, rbf_kernel, poly_kernel -
kernel.pars untyped - -
sgf.sv integer 5 [0,∞)[0, \infty)
sigf integer 7 [0,∞)[0, \infty)
maxiter integer 20 [0,∞)[0, \infty)
margin numeric 0.05 [0,∞)[0, \infty)
bound numeric 10 [0,∞)[0, \infty)
eig.tol numeric 1e-06 [0,∞)[0, \infty)
conv.tol numeric 1e-07 [0,∞)[0, \infty)
posd.tol numeric 1e-08 [0,∞)[0, \infty)

Prediction types

This learner returns up to two prediction types:

  1. crank: a vector containing the continuous ranking scores, where each score corresponds to a specific test observation.

  2. response: the survival time of each test observation, equal to -crank. This prediction type if only available for "type" equal to regression or hybrid.

Custom mlr3 parameters

  • gamma, mu have replaced gamma.mu so that it's easier to tune these separately. mu is only used when type = "hybrid".

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvSVM

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvSVM$new()

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvSVM$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Author(s)

RaphaelS1

References

Van Belle, Vanya, Pelckmans, Kristiaan, Van Huffel, Sabine, Suykens, AK J (2011). “Improved performance on high-dimensional survival data by application of Survival-SVM.” Bioinformatics, 27(1), 87–94.

Van Belle, Vanya, Pelckmans, Kristiaan, Van Huffel, Sabine, Suykens, AK J (2011). “Support vector methods for survival analysis: a comparison between ranking and regression approaches.” Artificial intelligence in medicine, 53(2), 107–118.

Shivaswamy, K P, Chu, Wei, Jansche, Martin (2007). “A support vector approach to censored targets.” In Seventh IEEE international conference on data mining (ICDM 2007), 655–660. IEEE.

See Also

Examples

set.seed(123)
# Define the Learner and set parameter values
learner = lrn("surv.svm", gamma = 0.1)
print(learner)

# Define a Task
task = tsk("rats")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

# print the model
print(learner$model)

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Extreme Gradient Boosting AFT Survival Learner

Description

eXtreme Gradient Boosting regression using an Accelerated Failure Time objective. Calls xgboost::xgb.train() from package xgboost with objective set to survival:aft and eval_metric to aft-nloglik.

Prediction types

This learner returns three prediction types:

  1. response: the estimated survival time TT for each test observation.

  2. lp: a vector of linear predictors (relative risk scores), one per observation, estimated as −log(T)-log(T). Higher survival time denotes lower risk.

  3. crank: same as lp.

Early stopping

Early stopping can be used to find the optimal number of boosting rounds. The early_stopping_set parameter controls which set is used to monitor the performance. By default, early_stopping_set = "none" which disables early stopping. Set early_stopping_set = "test" to monitor the performance of the model on the test set while training. The test set for early stopping can be set with the "test" row role in the mlr3::Task. Additionally, the range must be set in which the performance must increase with early_stopping_rounds and the maximum number of boosting rounds with nrounds. While resampling, the test set is automatically applied from the mlr3::Resampling. Not that using the test set for early stopping can potentially bias the performance scores.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.xgboost.aft")

Meta Information

Parameters

Id Type Default Levels Range
aft_loss_distribution character normal normal, logistic, extreme -
aft_loss_distribution_scale numeric - (−∞,∞)(-\infty, \infty)
alpha numeric 0 [0,∞)[0, \infty)
base_score numeric 0.5 (−∞,∞)(-\infty, \infty)
booster character gbtree gbtree, gblinear, dart -
callbacks untyped NULL -
seed integer - (−∞,∞)(-\infty, \infty)
colsample_bylevel numeric 1 [0,1][0, 1]
colsample_bynode numeric 1 [0,1][0, 1]
colsample_bytree numeric 1 [0,1][0, 1]
disable_default_eval_metric logical FALSE TRUE, FALSE -
evals untyped NULL -
early_stopping_rounds integer NULL [1,∞)[1, \infty)
learning_rate numeric 0.3 [0,1][0, 1]
feature_selector character cyclic cyclic, shuffle, random, greedy, thrifty -
gamma numeric 0 [0,∞)[0, \infty)
grow_policy character depthwise depthwise, lossguide -
interaction_constraints untyped - -
iterationrange untyped - -
lambda numeric 1 [0,∞)[0, \infty)
max_bin integer 256 [2,∞)[2, \infty)
max_delta_step numeric 0 [0,∞)[0, \infty)
max_depth integer 6 [0,∞)[0, \infty)
max_leaves integer 0 [0,∞)[0, \infty)
maximize logical NULL TRUE, FALSE -
min_child_weight numeric 1 [0,∞)[0, \infty)
monotone_constraints integer 0 [−1,1][-1, 1]
normalize_type character tree tree, forest -
nrounds integer - [1,∞)[1, \infty)
nthread integer - [1,∞)[1, \infty)
num_parallel_tree integer 1 [1,∞)[1, \infty)
one_drop logical FALSE TRUE, FALSE -
print_every_n integer 1 [1,∞)[1, \infty)
rate_drop numeric 0 [0,1][0, 1]
refresh_leaf logical TRUE TRUE, FALSE -
sampling_method character uniform uniform, gradient_based -
sample_type character uniform uniform, weighted -
save_name untyped - -
save_period integer - [0,∞)[0, \infty)
scale_pos_weight numeric 1 (−∞,∞)(-\infty, \infty)
seed_per_iteration logical FALSE TRUE, FALSE -
skip_drop numeric 0 [0,1][0, 1]
use_rmm logical FALSE TRUE, FALSE -
max_cached_hist_node integer NULL (−∞,∞)(-\infty, \infty)
max_cat_to_onehot integer 4 (−∞,∞)(-\infty, \infty)
max_cat_threshold integer 64 (−∞,∞)(-\infty, \infty)
subsample numeric 1 [0,1][0, 1]
top_k integer 0 [0,∞)[0, \infty)
tree_method character auto auto, exact, approx, hist, gpu_hist -
updater untyped - -
verbose integer - [0,2][0, 2]
verbosity integer - [0,2][0, 2]
xgb_model untyped - -
device untyped - -
missing numeric NA (−∞,∞)(-\infty, \infty)
validate_features logical TRUE TRUE, FALSE -

Initial parameter values

  • nrounds is initialized to 1000.

  • nthread is initialized to 1 to avoid conflicts with parallelization via future.

  • verbose and verbosity are both initialized to 0.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvXgboostAFT

Active bindings

internal_valid_scores

The last observation of the validation scores for all metrics. Extracted from model$evaluation_log

internal_tuned_values

Returns the early stopped iterations if early_stopping_rounds was set during training.

validate

How to construct the internal validation data. This parameter can be either NULL, a ratio, "test", or "predefined".

model

(any)
The fitted model. Only available after ⁠$train()⁠ has been called.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvXgboostAFT$new()

Method importance()

The importance scores are calculated with xgboost::xgb.importance().

Usage
LearnerSurvXgboostAFT$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvXgboostAFT$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

To compute on GPUs, you first need to compile xgboost yourself and link against CUDA. See https://xgboost.readthedocs.io/en/stable/build.html#building-with-gpu-support.

Author(s)

bblodfon

References

Chen, Tianqi, Guestrin, Carlos (2016). “Xgboost: A scalable tree boosting system.” In Proceedings of the 22nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 785–794. ACM. doi:10.1145/2939672.2939785.

Avinash B, Hyunsu C, Toby H (2022). “Survival Regression with Accelerated Failure Time Model in XGBoost.” Journal of Computational and Graphical Statistics. ISSN 15372715, doi:10.1080/10618600.2022.2067548.

See Also

Examples

# Define the Learner
learner = lrn("surv.xgboost.aft")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()

Extreme Gradient Boosting Cox Survival Learner

Description

eXtreme Gradient Boosting regression using a Cox Proportional Hazards objective. Calls xgboost::xgb.train() from package xgboost with objective set to survival:cox and eval_metric to cox-nloglik.

Prediction types

Three types of prediction are returned for this learner:

  1. lp: a vector of linear predictors (relative risk scores), one per observation.

  2. crank: same as lp.

  3. distr: a survival matrix in two dimensions, where observations are represented in rows and time points in columns. By default, the Breslow estimator is used via mlr3proba::breslow().

Initial parameter values

  • nrounds is initialized to 1000.

  • nthread is initialized to 1 to avoid conflicts with parallelization via future.

  • verbose and verbosity are both initialized to 0.

Dictionary

This Learner can be instantiated via lrn(): 

lrn("surv.xgboost.cox")

Meta Information

Parameters

Id Type Default Levels Range
alpha numeric 0 [0,∞)[0, \infty)
base_score numeric 0.5 (−∞,∞)(-\infty, \infty)
booster character gbtree gbtree, gblinear, dart -
callbacks untyped list() -
seed integer - (−∞,∞)(-\infty, \infty)
colsample_bylevel numeric 1 [0,1][0, 1]
colsample_bynode numeric 1 [0,1][0, 1]
colsample_bytree numeric 1 [0,1][0, 1]
disable_default_eval_metric logical FALSE TRUE, FALSE -
early_stopping_rounds integer NULL [1,∞)[1, \infty)
evals untyped NULL -
learning_rate numeric 0.3 [0,1][0, 1]
feature_selector character cyclic cyclic, shuffle, random, greedy, thrifty -
gamma numeric 0 [0,∞)[0, \infty)
grow_policy character depthwise depthwise, lossguide -
interaction_constraints untyped - -
iterationrange untyped - -
lambda numeric 1 [0,∞)[0, \infty)
max_bin integer 256 [2,∞)[2, \infty)
max_delta_step numeric 0 [0,∞)[0, \infty)
max_depth integer 6 [0,∞)[0, \infty)
max_leaves integer 0 [0,∞)[0, \infty)
maximize logical NULL TRUE, FALSE -
min_child_weight numeric 1 [0,∞)[0, \infty)
monotone_constraints integer 0 [−1,1][-1, 1]
normalize_type character tree tree, forest -
nrounds integer - [1,∞)[1, \infty)
nthread integer - [1,∞)[1, \infty)
num_parallel_tree integer 1 [1,∞)[1, \infty)
one_drop logical FALSE TRUE, FALSE -
print_every_n integer 1 [1,∞)[1, \infty)
rate_drop numeric 0 [0,1][0, 1]
refresh_leaf logical TRUE TRUE, FALSE -
sampling_method character uniform uniform, gradient_based -
sample_type character uniform uniform, weighted -
save_name untyped - -
save_period integer - [0,∞)[0, \infty)
scale_pos_weight numeric 1 (−∞,∞)(-\infty, \infty)
seed_per_iteration logical FALSE TRUE, FALSE -
skip_drop numeric 0 [0,1][0, 1]
use_rmm logical FALSE TRUE, FALSE -
max_cached_hist_node integer NULL (−∞,∞)(-\infty, \infty)
max_cat_to_onehot integer 4 (−∞,∞)(-\infty, \infty)
max_cat_threshold integer 64 (−∞,∞)(-\infty, \infty)
subsample numeric 1 [0,1][0, 1]
top_k integer 0 [0,∞)[0, \infty)
tree_method character auto auto, exact, approx, hist, gpu_hist -
updater untyped - -
verbose integer - [0,2][0, 2]
verbosity integer - [0,2][0, 2]
xgb_model untyped - -
device untyped - -
missing numeric NA (−∞,∞)(-\infty, \infty)
validate_features logical TRUE TRUE, FALSE -

Early stopping

Early stopping can be used to find the optimal number of boosting rounds. The early_stopping_set parameter controls which set is used to monitor the performance. By default, early_stopping_set = "none" which disables early stopping. Set early_stopping_set = "test" to monitor the performance of the model on the test set while training. The test set for early stopping can be set with the "test" row role in the mlr3::Task. Additionally, the range must be set in which the performance must increase with early_stopping_rounds and the maximum number of boosting rounds with nrounds. While resampling, the test set is automatically applied from the mlr3::Resampling. Not that using the test set for early stopping can potentially bias the performance scores.

Super classes

mlr3::Learner -> mlr3proba::LearnerSurv -> LearnerSurvXgboostCox

Active bindings

internal_valid_scores

(named list() or NULL) Validation metrics extracted from the xgboost model's evaluation_log. If early stopping is enabled, the scores correspond to the best_iteration selected by early stopping. Otherwise, the scores are taken from the final boosting round (nrounds).

internal_tuned_values

(named list() or NULL) If early stopping is activated, this returns a list with the early stopped iterations (nrounds), which is extracted from the best iteration of the model. Otherwise NULL.

validate

How to construct the internal validation data. This parameter can be either NULL, a ratio, "test", or "predefined".

model

(any)
The fitted model. Only available after ⁠$train()⁠ has been called.

Methods

Public methods

Inherited methods

Method new()

Creates a new instance of this R6 class.

Usage
LearnerSurvXgboostCox$new()

Method importance()

The importance scores are calculated with xgboost::xgb.importance().

Usage
LearnerSurvXgboostCox$importance()
Returns

Named numeric().

Method clone()

The objects of this class are cloneable with this method.

Usage
LearnerSurvXgboostCox$clone(deep = FALSE)
Arguments
deep

Whether to make a deep clone.

Note

To compute on GPUs, you first need to compile xgboost yourself and link against CUDA. See https://xgboost.readthedocs.io/en/stable/build.html#building-with-gpu-support.

Author(s)

bblodfon

References

Chen, Tianqi, Guestrin, Carlos (2016). “Xgboost: A scalable tree boosting system.” In Proceedings of the 22nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining, 785–794. ACM. doi:10.1145/2939672.2939785.

See Also

Examples

# Define the Learner
learner = lrn("surv.xgboost.cox")
print(learner)

# Define a Task
task = tsk("grace")

# Create train and test set
ids = partition(task)

# Train the learner on the training ids
learner$train(task, row_ids = ids$train)

print(learner$model)
print(learner$importance())

# Make predictions for the test rows
predictions = learner$predict(task, row_ids = ids$test)

# Score the predictions
predictions$score()