Changelog
mlr3extralearners 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
mlr3extralearners 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”
mlr3extralearners 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
mlr3extralearners 0.5.41
Fix bugs in learners lmer and J48
Remove predict type proba from J48
Delay loading of mlr3proba learners
mlr3extralearners 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
mlr3extralearners 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
mlr3extralearners 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
mlr3extralearners 0.5.35
- Full installatio in workflow ‘test_selection’ (is faster than the previous approach, where selected packages were installed from CRAN)
mlr3extralearners 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.
mlr3extralearners 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
mlr3extralearners 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 = …”
mlr3extralearners 0.5.25
- Introduce parameter
early_stopping_split
for lightgbm learners - Tidy description of R package
- Udpate NEWS.md for previous releases
mlr3extralearners 0.5.21
- Update files for creation of new learner
- Fixes regarding
create_learner
- CI modifications
mlr3extralearners 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
mlr3extralearners 0.5.18
- Allow integer as feature types for RWeka learners
- Correction of RWeka tests
mlr3extralearners 0.5.7
- Introduced new custom hyperparameters for
randomForestSRC::rfsrc()
,partykit::cforest()
andobliqueRSF::ORSF()
to conveniently tune hyperparameters whose upper limit depends on data dimensions.
mlr3extralearners 0.5.6
- Fix learners requiring distr6. distr6 1.6.0 now forced and param6 added to suggests
mlr3extralearners 0.5.4
- Add
regr.gausspr
andclassif.gausspr
fromkernlab::gausspr
mlr3extralearners 0.5.3
- Fixed bugs in catboost for classification
- Removed factor feature types from catboost
- Added
install_catboost
to make installation from catboost simpler
mlr3extralearners 0.5.2
Fixed learner tests # mlr3extralearners 0.5.1
Fixes bug in
base
parameter of {bart} learners
mlr3extralearners 0.5.0
- Deprecated liblinear learners now removed
- Internal changes to ParamSet representation
- checkmate now imported
mlr3extralearners 0.4.7
- Moved
nnet
learners to mlr3learners
mlr3extralearners 0.4.0
- Added
LearnerRegrGam
andLearnerClassifGam
with keysregr.gam
andclassif.gam
from packagemgcv
.
mlr3extralearners 0.3.0
- Added
LearnerRegrLightGBM
andLearnerClassifLightGBM
with keysregr.lightgbm
andclassif.lightgbm
respectively. Copied from mlr3learners.lightgbm -
LearnerRegrLiblineaRX
andLearnerClassifLiblineaRX
deprecated in favour of only two learners (LearnerRegrLiblineaR
andLearnerClassLiblineaR
) 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.
mlr3extralearners 0.2.0
-
dist = "logistic"
has been removed fromsurv.parametric
as it is unclear what this was previously predicting. - Added
type = "tobit"
fordist = "gaussian"
so predictions can correspond withsurvival::survreg
. - Added
LearnerRegrGlm
with the unique keyregr.glm
from packagestats
, which allows users to change thefamily
hyperparameter when fitting generalized linear regression models. - Minor internal changes
- Removed
keeptrees
parameter fromclassif.bart
as this is forced internally - Fixed incorrect response and probability predictions in
classif.bart
- Added hyper-parameters to
classif.earth
andregr.earth
- Added
se
predict type toregr.earth
- Fixed predictions in
regr.knn
andclassif.knn
mlr3extralearners 0.1.3
-
mlr3proba
moved toSuggests
-
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
mlr3extralearners 0.1.1
- Patch for bugs in
surv
learners that were reversing the order ofcrank
, see this issue for full details: https://github.com/mlr-org/mlr3proba/issues/165 -
response
is no longer returned bysurv.mboost
,surv.blackboost
,surv.glmboost
,surv.gamboost
orsurv.parametric
- Bugfix in
surv.parametric
withph
form - Bugfix in
survivalmodels
learners which weren’t returningdistr
-
surv.coxboost
andsurv.coxboost_cv
can now only handleinteger
andnumeric
feature types, previous automated internal coercions were inconsistent with mlr3 design.