Changes in version 1.5.2 Other - Use CRAN version of survdistr. - Use mlr3cmprsk version 0.0.5. - Update crs parameters. Changes in version 1.5.1 Other - Skip fastai and botorch tests on Windows and macOS where the Python backends crash or time out. - Skip tabpfn tests until token work reliable again. - Skip blockForest tests on macOS where SE predictions fail sanity checks. - Skip h2o.glm classification tests on Windows due to Java NullPointerException. - Skip GPfit tests on Windows where they crash under R-devel. - Skip classif.aorsf sanity autotest due to inconsistent tie-breaking across predict types. - Skip surv.flexreg sanity autotest on Windows due to initial parameter estimation failure. Changes in version 1.5.0 New Features - New Learners: - LearnerCompRisksCoxboost - LearnerRegrGPfit - LearnerClassifMLP - LearnerClassifSaeDNN - LearnerClassifPlsdaCaret - LearnerSurvDNN - LearnerRegrH2ORandomForest - LearnerRegrH2OGLM - LearnerClassifH2OGLM - LearnerClassifH2OGBM - LearnerClassifH2ORandomForest - LearnerClassifH2ODeeplearning - LearnerRegrH2OGBM - LearnerRegrH2ODeeplearning - LearnerClassifLvq1 - LearnerRegrBotorchFullyBayesian - Added kernel and input/output transformations for LearnerRegrBotorchSingleTaskGP and LearnerRegrBotorchMixedSingleTaskGP. Breaking Changes - Moved LearnerSurvAkritas and LearnerSurvParametric to the attic. See https://github.com/mlr-org/mlr3extralearners/issues/549. Other - Updated Extending vignette to incorporate information about skipping tests and considerations for testing Python learners - survdistr is now on Suggests (used for constant interpolation of the Kaplan-Meier predictions of the partykit survival learners) - Updated mlr3proba (0.8.8), pls and xgboost to the most recent CRAN versions Changes in version 1.4.0 New Features - New Learners: - LearnerSurvGamCox - LearnerSurvFlexReg - LearnerSurvNCVsurv - LearnerRegrRRF - LearnerRegrPcr - LearnerRegrPlsr - LearnerRegrLaGP - LearnerRegrFrbs - LearnerRegrBcart - LearnerRegrBgp - LearnerRegrBgpllm - LearnerRegrBlm - LearnerRegrBtgp - LearnerRegrBtgpllm - LearnerRegrBtlm - LearnerRegrNCVReg - LearnerClassifDbnDNN - LearnerClassifNNTrain - LearnerClassifSparseLDA - LearnerClassifNCVreg Breaking Changes - lrn("surv.flexible") (LearnerSurvFlexible) was renamed to lrn("surv.flexsurvspline") (LearnerSurvFlexSpline) to properly reflect the wrapped train function (Royston/Parmar spline model). Other - CoxBoost is now on CRAN, so we removed it from Remotes - lrn("surv.flexsurvspline") predicts linear predictors using predict.flexsurvreg(). We were doing manually the same exact prediction, so no functionality was changed. - compatibility: xgboost 3.1.2.1 (survival learners) - parameter updates for regr.lmer/glmer learners - updates for randomForestSRC 3.5.0 (use.uno parameter) - performance improvement: use of data.table::fifelse (@m-muecke) Changes in version 1.3.1 - Update website to include citation information Changes in version 1.3.0 - Add formula and anc params to surv.flexible learner, as well as response predict type (mean survival time). - Fix regr.gamboost regression predictions (#498). Changes in version 1.2.0 New Features - New Learners: - LearnerCompRisksRandomForestSRC - LearnerSurvBlockForest - Learner{Classif,Regr,Surv}BlockForest - Learner{Classif,Regr}ExhaustiveSearch - LearnerClassifFastai - Learner{Classif,Regr}Penalized - Learner{Classif,Regr}Bst - LearnerClassifAdabag - LearnerClassifAdaBoosting - Learner{Classif,Regr}Evtree - LearnerClassifKnn - LearnerClassifRotationForest - LearnerRegrCrs - LearnerClassifStepPlr - LearnerClassifMda - LearnerClassifRferns - LearnerClassifNeuralnet - LearnerRegrBrnn - LearnerRegrBotorchSingleTaskGP - LearnerRegrBotorchMixedSingleTaskGP - Add new control_custom_fun parameter in surv.aorsf - New function learner_is_runnable() to check whether the required packages to train a learner are available. - Added selected_features property to RandomForestSRC learners (prediction doesn't work if vars.used = 'all.trees') Bug fixes - Tests are now skipped when the suggested packages is not available. This will make local development much more convenient. - Removed parameters from RandomForestSRC learners that weren't used + optimized tests - Removed discrete parameter from surv.parametric, so that it is impossible to return distr6::VectorDistribution survival predictions (softly deprecated in mlr3proba@v0.8.1) Breaking Changes - All (extra) density learners are removed. These will be transferred to mlr3proba soon (see v0.8.2 or later). - The create_learner() generator was removed, because it was hard to maintain and boilerplate code in the age of LLMs is easy enough to write. - remove discrete parameter from surv.parametric, so that it is impossible to return distr6::VectorDistribution survival predictions (softly deprecated in mlr3proba@v0.8.1) - classif.lightgbm now works with encapsulation with multiclass tasks - the package no longer re-exports lrn and lrns, which should anyway be available to the user as the package depends on mlr3, where these functions are defined. - Removed various learners: - randomPlantedForest was removed, because there is currently no way to save the model. - The deep learning methods from survivalmodels were removed, because they also cannot be saved and because the upstream package is archived. Other - The package now imports withr - mlr3proba is now an import and no longer a suggested package. - mlr3cmprsk is added as an import. - The package no longer uses set.seed() in the tests and instead uses withr::local_seed() This means the auto tests will be stochastic like they should be. - The CI now checks that RCMD-check passes when suggested packages are not available. - distr6 dependency is removed. partykit survival learners use constant interpolation of the predicted Kaplan-Meier curves via survdistr::vec_interp() Changes in version 1.1.0 New Features: - Support offset in learners regr|classif.mgcv, regr.glm and regr.lmer. - Added learners LearnerRegrQGam and LearnerRegrMQGam. - Added learners LearnerClassifTabPFN and LearnerRegrTabPFN. - Added the new version of learner weights to all learners that support weights - Added marshaling for surv.xgboost.cox. - Added learner LearnerClassifKnn. Bugfixes: - lightgbm classifier now works with encapsulation (#437) Changes in version 1.0.0 - Add "Prediction types" doc section for all 30 survival learners + make sure it is consistent #347 - All survival learners have crank as main prediction type (and it is always returned) #331 - Added minimum working version for all survival learners in DESCRIPTION file - Harmonized the use of times points for prediction as much as possible across survival learners #387 - added gridify_times() function to coarse time points - fixed surv.parametric and surv.akritas use of ntime argument - surv.parametric is now used by default with discrete = TRUE (no survival learner returns now distr6 vectorized distribution by default) - Doc update for mlr3 (version 0.21.0) - Fixed custom and initial values across all learners documentation pages - Fixed doc examples that used learner$importance() - Set n_thread = 1 for surv.aorsf and use unique event time points for predicted S(t) - Add selected_features() for surv.penalized - Fix surv.prioritylasso learner + add distr predictions via Breslow #344 - Survival SVM gamma.mu parameter was split to gamma and mu to enable easier tuning (surv.svm learner) Changes in version 0.9.0 - Added response (i.e., survival time) prediction to aorsf learner - Updated support for flexsurv v2.3 - Fixed bug in catboost that caused invalid probability levels during resample() or benchmark() (#353) - the $model slot of lrn("classif.abess") now contains the model of the upstream package again. - Add early stopping and validation support to learners lrn("surv.xgboost.aft") and lrn("surv.xgboost.cox"). - Added early stopping and validation to catboost and lightgbm. - Added missing case.depth parameter to rfsrc learners. - mlr3 is now in Depends instead of Imports. - Deprecated learner lrn("surv.xgboost") was now removed. Use lrn("surv.xgboost.cox") or lrn("surv.xgboost.aft") instead. - Change xgboost default nrounds from 1 to 1000. - remove obliqueRSF Learner which was long superseded by aorsf - a lot of examples were added to the learners Changes in version 0.8.0 - Added surv.xgboost.cox and surv.xgboost.aft separate survival learners. distr prediction on the cox xgboost learner is now estimated via Breslow by default and aft xgboost has now in addition a response prediction (survival time) - Ported surv.parametric code to survivalmodels, changed type parameter to form to avoid conflict with survivalmodels's default parameter list - Fix: Replace hardcoded VectorDistributions from partykit and flexsurv survival learners with survival matrices (Matdist) (thanks to @bblodfon) - Feat: Add discrete parameter in surv.parametric learner to return Matdist survival predictions - Added method selected_features() to CoxBoost survival learners (thanks to @bblodfon) - Added the Random Planted Forest Learner (thanks to @jemus42) - re-added the catboost learner as it was requested (was previously removed because of installation issues) - surv.ranger now receives parameters during $predict() (thanks to @jemus42) - Feature: Learner surv.bart was added (thanks to @bblodfon) - Parameters of lrn("surv.aorsf") were updated (thanks to @bcjaeger) - Various minor doc improvements - Added the distr predict type to the surv.cv_glmnet and surv.glmnet learners (thanks to @bblodfon) - Feat: Added many new WEKA learners (thanks to @damirpolat) - Fix: I and F params from IBk learner are too interdependent (I can only be TRUE when F is FALSE and vice versa). Combined them into one factor param weight that has two levels -- I and F. - Fix: U must be FALSE for S to be tunable in J48 learner. - Compatibility with upcoming 'paradox' release. Changes in version 0.7.1 - Add parameter perf.type to rfsrc learners - Add vignette about "extending learners" which was previously in the mlr3book. - Remove the "multiclass" property from lrn("classif.gbm"), as this feature is broken. Changes in version 0.7.0 - Add new parameters to lightgbm learners - Add feature type "factor" to gam learners - Add new parameter min.bucket to ranger - Remove catboost learner (because the developers don't properly take care of the R package) - Add argument nthreads to dbarts learners; set verbose to FALSE by default (thanks to @ck37) - Add new parameters to prioritylasso - Fix: available levels for parameter of imbalanced random forest (typo) Changes in version 0.6.1 - BREAKING CHANGE: lightgbm's early stopping mechanism now uses the task's test set. - feat: Add two new learners regr.abess and classif.abess (thanks to @bbayukari) - feat: Added learner LearnerClassifImbalancedRandomForestSRC (thanks to @HarutyunyanLiana) - Feat: Added learners LearnerClassifPriorityLasso, LearnerRegrPriorityLasso, LearnerSurvPriorityLasso (thanks to @HarutyunyanLiana) Changes in version 0.6.0 - Feat: Added learner LearnerClassifGlmer (https://github.com/mlr-org/mlr3extralearners/issues/243) - Fix: Failing xgboost parameter test - Fix: Add arguments nei and ncv.thread that were added to mgcv::gam() in version 1.8.41 - Fix: Added missing property "weights" to LearnerClassifGlmer and LearnerRegrLmer - Fix: lightgbm uses the param_vals stored in the state for hotstarting - Fix: Rely on state$data_prototype to get ordering of features via ordered_features() like in mlr3learners and therefore obviate the need to store feature_names in the state - Fix: extralearners are removed from mlr_learners when unloading mlr3extralearners Changes in version 0.5.49 - Added missing feature type "integer" to classif.randomForest - Added missing feature type "logical" to {classif, regr}.randomForest Changes in version 0.5.48 - Add rsm learner - fix list_mlr3learners() function. Now slower but correct. - Remove catboost from DESCRIPTION until it can be installed with pak - Fix typos in test templates - Update README Changes in version 0.5.47 - Add mlr3proba dependencies into remotes (no longer on CRAN) - Correct documentation of gbm learner: default was incorrectly documented and the parameter was incorrectly referred to as keep_data instead of keep.data - Add catboost to the dependencies - Added LearnerSurvAorsf with key surv.aorsf. See https://github.com/bcjaeger/aorsf for more details on aorsf Changes in version 0.5.46 - Addresses https://github.com/mlr-org/mlr3extralearners/issues/225 - Fix link in README - Fix learner status overview Changes in version 0.5.45 - Minor corrections in create_learner and the learner template. Changes in version 0.5.44 - Corrected parameters in lightgbm learners - Implemented hotstarting for lightgbm learners - Adjusted lightgbm train and predict methods to changes in lightgbm dev version (https://github.com/mlr-org/mlr3extralearners/issues/217) - Added paramtests for lightgbm through webscraping Changes in version 0.5.43 - Clean up test files - Fix installation of catboost in CI - Fix the create_learner function - Adjust templates for creation of learner - Split up "Parameter Changes" in sections "Custom mlr3 parameters" and "Custom mlr3 defaults" Changes in version 0.5.42 - Fix bug in C50 learner: Weights were not passed correctly - Remove kerdiest Learner because it is not being maintained on CRAN anymore Changes in version 0.5.41 - Fix bugs in learners lmer and J48 - Remove predict type proba from J48 - Delay loading of mlr3proba learners Changes in version 0.5.40 - lightgbm: - Add parameter convert_categoricals - Validation split not respects grouping / stratification - Fixed bug - Docs: Renamed section "Custom mlr3 defaults" to "Parameter Changes" - Added labels to learners Changes in version 0.5.39 - Remove extraTrees because it is no longer on CRAN and GH version has errors - Remove sketch_eps parameter from xgboost because it is no longer listed in the docs Changes in version 0.5.38 - Added regr.lmer Changes in version 0.5.37 - Improve docs and change doc layout - Fix typo in man-roxygen templates - Port mlr3proba learners (mlr3proba is no longer on CRAN) - Exclude relevant files in precommit Changes in version 0.5.36 - Add missing 'threads' tag to respective parameters. Changes in version 0.5.35 - Full installatio in workflow 'test_selection' (is faster than the previous approach, where selected packages were installed from CRAN) Changes in version 0.5.34 - remove explicit mlr3misc:: (is imported) Changes in version 0.5.33 - consistency: Use params in train and predict calls, even in learners that currently don't have predict / train params. This allows easier correction of parameters by users. Changes in version 0.5.32 - chore: add new parameters for kde and rfsrc - temporarily disable feat_all test for obliqeRSF (failed in $score() stage, because issue only happened in CI and could not be reproduced Changes in version 0.5.31 - Many non-standard tags were included in the learners, these are removed - Some bugs in learners were fixed (survival rfsrc: "estimator" was incorrectly handled in .predict) - Minor refactorings in train methods of learners - Avoid partial argument matching: Some learners used "tag = ..." instead of the correct "tags = ..." Changes in version 0.5.30 - Revert to using mlr3proba and survivalmodels CRAN version Changes in version 0.5.29 - Change in vignette Changes in version 0.5.28 - update randomForestSRC Changes in version 0.5.27 - Update learner status page Changes in version 0.5.26 - Fixed survivalmodel learners Changes in version 0.5.25 - Introduce parameter early_stopping_split for lightgbm learners - Tidy description of R package - Udpate NEWS.md for previous releases Changes in version 0.5.24 - Don't allow integer for density estimator dens.plug Changes in version 0.5.23 - Fix bug in lightgbm Changes in version 0.5.22 - Style package using the mlr3 style Changes in version 0.5.21 - Update files for creation of new learner - Fixes regarding create_learner - CI modifications Changes in version 0.5.20 - Fix all parameter tests (run_paramtest was updated in mlr3 in November 2021) - paramtests were moved from inst/paramtest to tests/testthat - Change in the CI files: parameter tests and learner tests are now run together - formatting and other minor corrections Changes in version 0.5.19 - Provide correct range for neighors argument for Cubist Changes in version 0.5.18 - Allow integer as feature types for RWeka learners - Correction of RWeka tests Changes in version 0.5.17 - Improve vignette Changes in version 0.5.16 - Fix bug in AdaBoostM1 (control arg) Changes in version 0.5.15 - Change in maintainer Changes in version 0.5.14 - Fix bug regarding Weka control args. Changes in version 0.5.13 - Fix categorical_features in {lightgbm} learners Changes in version 0.5.12 - Patch for lightgbm updates Changes in version 0.5.11 - Add option to not open files with create_learner Changes in version 0.5.10 - Added params ignored_features and one_hot_max_size to classif.catboost Changes in version 0.5.9 - Fixed bug that didn't allow C parameter to be set for nu-regression Changes in version 0.5.8 - Add regr.rvm and classif.lssvm Changes in version 0.5.7 - Introduced new custom hyperparameters for randomForestSRC::rfsrc(), partykit::cforest() and obliqueRSF::ORSF() to conveniently tune hyperparameters whose upper limit depends on data dimensions. Changes in version 0.5.6 - Fix learners requiring distr6. distr6 1.6.0 now forced and param6 added to suggests Changes in version 0.5.5 - Bugfix regr.gausspr Changes in version 0.5.4 - Add regr.gausspr and classif.gausspr from kernlab::gausspr Changes in version 0.5.3 - Fixed bugs in catboost for classification - Removed factor feature types from catboost - Added install_catboost to make installation from catboost simpler Changes in version 0.5.2 - Fixed learner tests Changes in version 0.5.1 - Fixes bug in base parameter of {bart} learners Changes in version 0.5.0 - Deprecated liblinear learners now removed - Internal changes to ParamSet representation - checkmate now imported Changes in version 0.4.9 - Minor internal changes Changes in version 0.4.8 - Added LearnerRegrCubist and LearnerRegrMars Changes in version 0.4.7 - Moved nnet learners to mlr3learners Changes in version 0.4.6 - Updates default cores for rfsrc learners to 1 Changes in version 0.4.5 - Fix RWeka tests (stochastic failures, implementation unaffected) Changes in version 0.4.3 - Add support for custom families in all remaining mboost learners Changes in version 0.4.2 - Fix broken partykit tests Changes in version 0.4.0 - Added LearnerRegrGam and LearnerClassifGam with keys regr.gam and classif.gam from package mgcv. Changes in version 0.3.6 - surv.coxboost now uses the GitHub version instead of CRAN (archived) Changes in version 0.3.4 - Add support for custom families to regr.glmboost Changes in version 0.3.1 - surv.svm now supports all feature types Changes in version 0.3.0 - Added LearnerRegrLightGBM and LearnerClassifLightGBM with keys regr.lightgbm and classif.lightgbm respectively. Copied from mlr3learners.lightgbm - LearnerRegrLiblineaRX and LearnerClassifLiblineaRX deprecated in favour of only two learners (LearnerRegrLiblineaR and LearnerClassLiblineaR) with added hyper-parameters. Deprecated learners will be removed in v0.3.0. - Deprecated classif.nnet will be removed in v0.4.0. - Deprecated liblinearX will be removed in v0.4.0. Changes in version 0.2.0 - dist = "logistic" has been removed from surv.parametric as it is unclear what this was previously predicting. - Added type = "tobit" for dist = "gaussian" so predictions can correspond with survival::survreg. - Added LearnerRegrGlm with the unique key regr.glm from package stats, which allows users to change the family hyperparameter when fitting generalized linear regression models. - Minor internal changes - Removed keeptrees parameter from classif.bart as this is forced internally - Fixed incorrect response and probability predictions in classif.bart - Added hyper-parameters to classif.earth and regr.earth - Added se predict type to regr.earth - Fixed predictions in regr.knn and classif.knn Changes in version 0.1.3 - mlr3proba moved to Suggests - install_learners now additionally installs required mlr3 packages - Bugfix in surv.parametric occurring if feature names are switched between training and predicting - Deprecated classif.nnet, in the future please load from mlr3learners Changes in version 0.1.2 - Fixes in crank and distr computation of all survival learners Changes in version 0.1.1 - Patch for bugs in surv learners that were reversing the order of crank, see this issue for full details: https://github.com/mlr-org/mlr3proba/issues/165 - response is no longer returned by surv.mboost, surv.blackboost, surv.glmboost, surv.gamboost or surv.parametric - Bugfix in surv.parametric with ph form - Bugfix in survivalmodelslearners which weren't returning distr - surv.coxboost and surv.coxboost_cv can now only handle integer and numeric feature types, previous automated internal coercions were inconsistent with mlr3 design. Changes in version 0.1.0 - Initial release. mlr3extralearners contains all learners from the mlr3learners organisation, which is now archived.