- Other changes:
- Updated
ndarrayto 0.17,randto 0.9 andrayonto 1.11. Thanks @Havunen!
- Updated
- Other changes:
- Update
maturin.
- Update
- Other changes:
- Updated
ndarrayto 0.16. Thanks @Havunen!
- Updated
-
Breaking changes:
- The arguments
mtryandseedtoDecisionTreeandRandomForesthave been renamed tomax_featuresandrandom_state, aligning them with their scikit-learn counterparts. - Supplying
n_jobs=Nonewill now result in no parallelization, aligning its behaviour with scikit-learn. To use all processes, usen_jobs=-1.
- The arguments
-
New features:
- The
max_featuresparameter for classesDecisionTreeandRandomForestcan now be supplied with a fraction, an integer,Noneand"sqrt".
- The
- Other changes:
- Speedup of
DecisisionTreeNode.split_samplesresulting in overall 6 - 20% faster tree fitting.
- Speedup of
- Bug fixes:
DecisionTreeNodeno longer returns wrong leaf value if splitting is stopped due tomin_samples_split.
New features:
- Parallization for RandomForest::fit and RandomForest::fit_predict_oob.