doc/whats_new/v0.23.rst
.. include:: _contributors.rst
.. currentmodule:: sklearn
.. _release_notes_0_23:
For a short description of the main highlights of the release, please refer to
:ref:sphx_glr_auto_examples_release_highlights_plot_release_highlights_0_23_0.py.
.. include:: changelog_legend.inc
.. _changes_0_23_2:
The following estimators and functions, when fit with the same data and parameters, may produce different models from the previous version. This often occurs due to changes in the modelling logic (bug fixes or enhancements), or in random sampling procedures.
inertia_ attribute of :class:cluster.KMeans and
:class:cluster.MiniBatchKMeans.Details are listed in the changelog below.
(While we are trying to better inform users by providing this information, we cannot assure that this list is complete.)
:mod:sklearn.cluster
......................
|Fix| Fixed a bug in :class:cluster.KMeans where rounding errors could
prevent convergence to be declared when tol=0. :pr:17959 by
:user:Jérémie du Boisberranger <jeremiedbb>.
|Fix| Fixed a bug in :class:cluster.KMeans and
:class:cluster.MiniBatchKMeans where the reported inertia was incorrectly
weighted by the sample weights. :pr:17848 by
:user:Jérémie du Boisberranger <jeremiedbb>.
|Fix| Fixed a bug in :class:cluster.MeanShift with bin_seeding=True. When
the estimated bandwidth is 0, the behavior is equivalent to
bin_seeding=False.
:pr:17742 by :user:Jeremie du Boisberranger <jeremiedbb>.
|Fix| Fixed a bug in :class:cluster.AffinityPropagation, that
gives incorrect clusters when the array dtype is float32.
:pr:17995 by :user:Thomaz Santana <Wikilicious> and
:user:Amanda Dsouza <amy12xx>.
:mod:sklearn.decomposition
............................
|Fix| Fixed a bug in
:func:decomposition.MiniBatchDictionaryLearning.partial_fit which should
update the dictionary by iterating only once over a mini-batch.
:pr:17433 by :user:Chiara Marmo <cmarmo>.
|Fix| Avoid overflows on Windows in
:func:decomposition.IncrementalPCA.partial_fit for large batch_size and
n_samples values.
:pr:17985 by :user:Alan Butler <aldee153> and
:user:Amanda Dsouza <amy12xx>.
:mod:sklearn.ensemble
.......................
|Fix| Fixed bug in ensemble.MultinomialDeviance where the
average of logloss was incorrectly calculated as sum of logloss.
:pr:17694 by :user:Markus Rempfler <rempfler> and
:user:Tsutomu Kusanagi <t-kusanagi2>.
|Fix| Fixes :class:ensemble.StackingClassifier and
:class:ensemble.StackingRegressor compatibility with estimators that
do not define n_features_in_. :pr:17357 by Thomas Fan_.
:mod:sklearn.feature_extraction
.................................
feature_extraction.text.CountVectorizer where
sample order invariance was broken when max_features was set and features
had the same count. :pr:18016 by Thomas Fan, Roman Yurchak, and
Joel Nothman_.:mod:sklearn.linear_model
...........................
linear_model.lars_path does not overwrite X when
X_copy=True and Gram='auto'. :pr:17914 by Thomas Fan_.:mod:sklearn.manifold
.......................
metrics.pairwise_distances would raise an
error if metric='seuclidean' and X is not type np.float64.
:pr:15730 by :user:Forrest Koch <ForrestCKoch>.:mod:sklearn.metrics
......................
metrics.mean_squared_error where the
average of multiple RMSE values was incorrectly calculated as the root of the
average of multiple MSE values.
:pr:17309 by :user:Swier Heeres <swierh>.:mod:sklearn.pipeline
.......................
pipeline.FeatureUnion raises a deprecation warning when
None is included in transformer_list. :pr:17360 by Thomas Fan_.:mod:sklearn.utils
....................
utils.estimator_checks.check_estimator so that all test
cases support the binary_only estimator tag.
:pr:17812 by :user:Bruno Charron <brcharron>... _changes_0_23_1:
May 18 2020
:mod:sklearn.cluster
......................
|Efficiency| :class:cluster.KMeans efficiency has been improved for very
small datasets. In particular it cannot spawn idle threads any more.
:pr:17210 and :pr:17235 by :user:Jeremie du Boisberranger <jeremiedbb>.
|Fix| Fixed a bug in :class:cluster.KMeans where the sample weights
provided by the user were modified in place. :pr:17204 by
:user:Jeremie du Boisberranger <jeremiedbb>.
Miscellaneous .............
repr of third-party estimators that use a
**kwargs parameter in their constructor, when changed_only is True
which is now the default. :pr:17205 by Nicolas Hug_... _changes_0_23:
May 12 2020
In an effort to promote clear and non-ambiguous use of the library, most
constructor and function parameters are now expected to be passed as keyword
arguments (i.e. using the param=value syntax) instead of positional. To
ease the transition, a FutureWarning is raised if a keyword-only parameter
is used as positional. In version 1.0 (renaming of 0.25), these parameters
will be strictly keyword-only, and a TypeError will be raised.
:issue:15005 by Joel Nothman, Adrin Jalali, Thomas Fan, and
Nicolas Hug. See SLEP009 <https://scikit-learn-enhancement-proposals.readthedocs.io/en/latest/slep009/proposal.html>_
for more details.
The following estimators and functions, when fit with the same data and parameters, may produce different models from the previous version. This often occurs due to changes in the modelling logic (bug fixes or enhancements), or in random sampling procedures.
ensemble.BaggingClassifier, :class:ensemble.BaggingRegressor,
and :class:ensemble.IsolationForest.cluster.KMeans with algorithm="elkan" and
algorithm="full".cluster.Birchcompose.ColumnTransformer.get_feature_namescompose.ColumnTransformer.fitdatasets.make_multilabel_classificationdecomposition.PCA with n_components='mle'decomposition.NMF and
:func:decomposition.non_negative_factorization with float32 dtype input.decomposition.KernelPCA.inverse_transformensemble.HistGradientBoostingClassifier and
:class:ensemble.HistGradientBoostingRegressorestimator_samples_ in :class:ensemble.BaggingClassifier,
:class:ensemble.BaggingRegressor and :class:ensemble.IsolationForestensemble.StackingClassifier and
:class:ensemble.StackingRegressor with sample_weightgaussian_process.GaussianProcessRegressorlinear_model.RANSACRegressor with sample_weight.linear_model.RidgeClassifierCVmetrics.mean_squared_error with squared and
multioutput='raw_values'.metrics.mutual_info_score with negative scores.metrics.confusion_matrix with zero length y_true and y_predneural_network.MLPClassifierpreprocessing.StandardScaler with partial_fit and sparse
input.preprocessing.Normalizer with norm='max'svm.libsvm or the svm.liblinear solver,
including :class:svm.LinearSVC, :class:svm.LinearSVR,
:class:svm.NuSVC, :class:svm.NuSVR, :class:svm.OneClassSVM,
:class:svm.SVC, :class:svm.SVR, :class:linear_model.LogisticRegression.tree.DecisionTreeClassifier, :class:tree.ExtraTreeClassifier and
:class:ensemble.GradientBoostingClassifier as well as predict method of
:class:tree.DecisionTreeRegressor, :class:tree.ExtraTreeRegressor, and
:class:ensemble.GradientBoostingRegressor and read-only float32 input in
predict, decision_path and predict_proba.Details are listed in the changelog below.
(While we are trying to better inform users by providing this information, we cannot assure that this list is complete.)
..
Entries should be grouped by module (in alphabetic order) and prefixed with
one of the labels: |MajorFeature|, |Feature|, |Efficiency|, |Enhancement|,
|Fix| or |API| (see whats_new.rst for descriptions).
Entries should be ordered by those labels (e.g. |Fix| after |Efficiency|).
Changes not specific to a module should be listed under Multiple Modules
or Miscellaneous.
Entries should end with:
:pr:123456 by :user:Joe Bloggs <joeongithub>.
where 123456 is the pull request number, not the issue number.
:mod:sklearn.cluster
......................
|Efficiency| :class:cluster.Birch implementation of the predict method
avoids high memory footprint by calculating the distances matrix using
a chunked scheme.
:pr:16149 by :user:Jeremie du Boisberranger <jeremiedbb> and
:user:Alex Shacked <alexshacked>.
|Efficiency| |MajorFeature| The critical parts of :class:cluster.KMeans
have a more optimized implementation. Parallelism is now over the data
instead of over initializations allowing better scalability. :pr:11950 by
:user:Jeremie du Boisberranger <jeremiedbb>.
|Enhancement| :class:cluster.KMeans now supports sparse data when
solver = "elkan". :pr:11950 by
:user:Jeremie du Boisberranger <jeremiedbb>.
|Enhancement| :class:cluster.AgglomerativeClustering has a faster and more
memory efficient implementation of single linkage clustering.
:pr:11514 by :user:Leland McInnes <lmcinnes>.
|Fix| :class:cluster.KMeans with algorithm="elkan" now converges with
tol=0 as with the default algorithm="full". :pr:16075 by
:user:Erich Schubert <kno10>.
|Fix| Fixed a bug in :class:cluster.Birch where the n_clusters parameter
could not have a np.int64 type. :pr:16484
by :user:Jeremie du Boisberranger <jeremiedbb>.
|Fix| :class:cluster.AgglomerativeClustering add specific error when
distance matrix is not square and affinity=precomputed.
:pr:16257 by :user:Simona Maggio <simonamaggio>.
|API| The n_jobs parameter of :class:cluster.KMeans,
:class:cluster.SpectralCoclustering and
:class:cluster.SpectralBiclustering is deprecated. They now use OpenMP
based parallelism. For more details on how to control the number of threads,
please refer to our :ref:parallelism notes. :pr:11950 by
:user:Jeremie du Boisberranger <jeremiedbb>.
|API| The precompute_distances parameter of :class:cluster.KMeans is
deprecated. It has no effect. :pr:11950 by
:user:Jeremie du Boisberranger <jeremiedbb>.
|API| The random_state parameter has been added to
:class:cluster.AffinityPropagation. :pr:16801 by :user:rcwoolston
and :user:Chiara Marmo <cmarmo>.
:mod:sklearn.compose
......................
|Efficiency| :class:compose.ColumnTransformer is now faster when working
with dataframes and strings are used to specific subsets of data for
transformers. :pr:16431 by Thomas Fan_.
|Enhancement| :class:compose.ColumnTransformer method get_feature_names
now supports 'passthrough' columns, with the feature name being either
the column name for a dataframe, or 'xi' for column index i.
:pr:14048 by :user:Lewis Ball <lrjball>.
|Fix| :class:compose.ColumnTransformer method get_feature_names now
returns correct results when one of the transformer steps applies on an
empty list of columns :pr:15963 by Roman Yurchak_.
|Fix| :func:compose.ColumnTransformer.fit will error when selecting
a column name that is not unique in the dataframe. :pr:16431 by
Thomas Fan_.
:mod:sklearn.datasets
.......................
|Efficiency| :func:datasets.fetch_openml has reduced memory usage because
it no longer stores the full dataset text stream in memory. :pr:16084 by
Joel Nothman_.
|Feature| :func:datasets.fetch_california_housing now supports
heterogeneous data using pandas by setting as_frame=True. :pr:15950
by :user:Stephanie Andrews <gitsteph> and
:user:Reshama Shaikh <reshamas>.
|Feature| embedded dataset loaders :func:datasets.load_breast_cancer,
:func:datasets.load_diabetes, :func:datasets.load_digits,
:func:datasets.load_iris, :func:datasets.load_linnerud and
:func:datasets.load_wine now support loading as a pandas DataFrame by
setting as_frame=True. :pr:15980 by :user:wconnell and
:user:Reshama Shaikh <reshamas>.
|Enhancement| Added return_centers parameter in
:func:datasets.make_blobs, which can be used to return
centers for each cluster.
:pr:15709 by :user:shivamgargsya and
:user:Venkatachalam N <venkyyuvy>.
|Enhancement| Functions :func:datasets.make_circles and
:func:datasets.make_moons now accept two-element tuple.
:pr:15707 by :user:Maciej J Mikulski <mjmikulski>.
|Fix| :func:datasets.make_multilabel_classification now generates
ValueError for arguments n_classes < 1 OR length < 1.
:pr:16006 by :user:Rushabh Vasani <rushabh-v>.
|API| The StreamHandler was removed from sklearn.logger to avoid
double logging of messages in common cases where a handler is attached
to the root logger, and to follow the Python logging documentation
recommendation for libraries to leave the log message handling to
users and application code. :pr:16451 by :user:Christoph Deil <cdeil>.
:mod:sklearn.decomposition
............................
|Enhancement| :class:decomposition.NMF and
:func:decomposition.non_negative_factorization now preserves float32 dtype.
:pr:16280 by :user:Jeremie du Boisberranger <jeremiedbb>.
|Enhancement| :func:decomposition.TruncatedSVD.transform is now faster on
given sparse csc matrices. :pr:16837 by :user:wornbb.
|Fix| :class:decomposition.PCA with a float n_components parameter, will
exclusively choose the components that explain the variance greater than
n_components. :pr:15669 by :user:Krishna Chaitanya <krishnachaitanya9>
|Fix| :class:decomposition.PCA with n_components='mle' now correctly
handles small eigenvalues, and does not infer 0 as the correct number of
components. :pr:16224 by :user:Lisa Schwetlick <lschwetlick>, and
:user:Gelavizh Ahmadi <gelavizh1> and :user:Marija Vlajic Wheeler <marijavlajic> and :pr:16841 by Nicolas Hug_.
|Fix| :class:decomposition.KernelPCA method inverse_transform now
applies the correct inverse transform to the transformed data. :pr:16655
by :user:Lewis Ball <lrjball>.
|Fix| Fixed bug that was causing :class:decomposition.KernelPCA to sometimes
raise invalid value encountered in multiply during fit.
:pr:16718 by :user:Gui Miotto <gui-miotto>.
|Feature| Added n_components_ attribute to :class:decomposition.SparsePCA
and :class:decomposition.MiniBatchSparsePCA. :pr:16981 by
:user:Mateusz Górski <Reksbril>.
:mod:sklearn.ensemble
.......................
|MajorFeature| :class:ensemble.HistGradientBoostingClassifier and
:class:ensemble.HistGradientBoostingRegressor now support
:term:sample_weight. :pr:14696 by Adrin Jalali_ and Nicolas Hug_.
|Feature| Early stopping in
:class:ensemble.HistGradientBoostingClassifier and
:class:ensemble.HistGradientBoostingRegressor is now determined with a
new early_stopping parameter instead of n_iter_no_change. Default value
is 'auto', which enables early stopping if there are at least 10,000
samples in the training set. :pr:14516 by :user:Johann Faouzi <johannfaouzi>.
|MajorFeature| :class:ensemble.HistGradientBoostingClassifier and
:class:ensemble.HistGradientBoostingRegressor now support monotonic
constraints, useful when features are supposed to have a positive/negative
effect on the target. :pr:15582 by Nicolas Hug_.
|API| Added boolean verbose flag to classes:
:class:ensemble.VotingClassifier and :class:ensemble.VotingRegressor.
:pr:16069 by :user:Sam Bail <spbail>,
:user:Hanna Bruce MacDonald <hannahbrucemacdonald>,
:user:Reshama Shaikh <reshamas>, and
:user:Chiara Marmo <cmarmo>.
|API| Fixed a bug in :class:ensemble.HistGradientBoostingClassifier and
:class:ensemble.HistGradientBoostingRegressor that would not respect the
max_leaf_nodes parameter if the criteria was reached at the same time as
the max_depth criteria. :pr:16183 by Nicolas Hug_.
|Fix| Changed the convention for max_depth parameter of
:class:ensemble.HistGradientBoostingClassifier and
:class:ensemble.HistGradientBoostingRegressor. The depth now corresponds to
the number of edges to go from the root to the deepest leaf.
Stumps (trees with one split) are now allowed.
:pr:16182 by :user:Santhosh B <santhoshbala18>
|Fix| Fixed a bug in :class:ensemble.BaggingClassifier,
:class:ensemble.BaggingRegressor and :class:ensemble.IsolationForest
where the attribute estimators_samples_ did not generate the proper indices
used during fit.
:pr:16437 by :user:Jin-Hwan CHO <chofchof>.
|Fix| Fixed a bug in :class:ensemble.StackingClassifier and
:class:ensemble.StackingRegressor where the sample_weight
argument was not being passed to cross_val_predict when
evaluating the base estimators on cross-validation folds
to obtain the input to the meta estimator.
:pr:16539 by :user:Bill DeRose <wderose>.
|Feature| Added additional option loss="poisson" to
:class:ensemble.HistGradientBoostingRegressor, which adds Poisson deviance
with log-link useful for modeling count data.
:pr:16692 by :user:Christian Lorentzen <lorentzenchr>
|Fix| Fixed a bug where :class:ensemble.HistGradientBoostingRegressor and
:class:ensemble.HistGradientBoostingClassifier would fail with multiple
calls to fit when warm_start=True, early_stopping=True, and there is no
validation set. :pr:16663 by Thomas Fan_.
:mod:sklearn.feature_extraction
.................................
feature_extraction.text.CountVectorizer now sorts
features after pruning them by document frequency. This improves performances
for datasets with large vocabularies combined with min_df or max_df.
:pr:15834 by :user:Santiago M. Mola <smola>.:mod:sklearn.feature_selection
................................
|Enhancement| Added support for multioutput data in
:class:feature_selection.RFE and :class:feature_selection.RFECV.
:pr:16103 by :user:Divyaprabha M <divyaprabha123>.
|API| Adds :class:feature_selection.SelectorMixin back to public API.
:pr:16132 by :user:trimeta.
:mod:sklearn.gaussian_process
...............................
|Enhancement| :func:gaussian_process.kernels.Matern returns the RBF kernel when nu=np.inf.
:pr:15503 by :user:Sam Dixon <sam-dixon>.
|Fix| Fixed bug in :class:gaussian_process.GaussianProcessRegressor that
caused predicted standard deviations to only be between 0 and 1 when
WhiteKernel is not used. :pr:15782
by :user:plgreenLIRU.
:mod:sklearn.impute
.....................
|Enhancement| :class:impute.IterativeImputer accepts both scalar and array-like inputs for
max_value and min_value. Array-like inputs allow a different max and min to be specified
for each feature. :pr:16403 by :user:Narendra Mukherjee <narendramukherjee>.
|Enhancement| :class:impute.SimpleImputer, :class:impute.KNNImputer, and
:class:impute.IterativeImputer accepts pandas' nullable integer dtype with
missing values. :pr:16508 by Thomas Fan_.
:mod:sklearn.inspection
.........................
inspection.partial_dependence and
inspection.plot_partial_dependence now support the fast 'recursion'
method for :class:ensemble.RandomForestRegressor and
:class:tree.DecisionTreeRegressor. :pr:15864 by
Nicolas Hug_.:mod:sklearn.linear_model
...........................
|MajorFeature| Added generalized linear models (GLM) with non normal error
distributions, including :class:linear_model.PoissonRegressor,
:class:linear_model.GammaRegressor and :class:linear_model.TweedieRegressor
which use Poisson, Gamma and Tweedie distributions respectively.
:pr:14300 by :user:Christian Lorentzen <lorentzenchr>, Roman Yurchak,
and Olivier Grisel.
|MajorFeature| Support of sample_weight in
:class:linear_model.ElasticNet and :class:linear_model.Lasso for dense
feature matrix X. :pr:15436 by :user:Christian Lorentzen <lorentzenchr>.
|Efficiency| :class:linear_model.RidgeCV and
:class:linear_model.RidgeClassifierCV now do not allocate a
potentially large array to store dual coefficients for all hyperparameters
during its fit, nor an array to store all error or LOO predictions unless
store_cv_values is True.
:pr:15652 by :user:Jérôme Dockès <jeromedockes>.
|Enhancement| :class:linear_model.LassoLars and
:class:linear_model.Lars now support a jitter parameter that adds
random noise to the target. This might help with stability in some edge
cases. :pr:15179 by :user:angelaambroz.
|Fix| Fixed a bug where if a sample_weight parameter was passed to the fit
method of :class:linear_model.RANSACRegressor, it would not be passed to
the wrapped base_estimator during the fitting of the final model.
:pr:15773 by :user:Jeremy Alexandre <J-A16>.
|Fix| Add best_score_ attribute to :class:linear_model.RidgeCV and
:class:linear_model.RidgeClassifierCV.
:pr:15655 by :user:Jérôme Dockès <jeromedockes>.
|Fix| Fixed a bug in :class:linear_model.RidgeClassifierCV to pass a
specific scoring strategy. Before the internal estimator outputs score
instead of predictions.
:pr:14848 by :user:Venkatachalam N <venkyyuvy>.
|Fix| :class:linear_model.LogisticRegression will now avoid an unnecessary
iteration when solver='newton-cg' by checking for inferior or equal instead
of strictly inferior for maximum of absgrad and tol in utils.optimize._newton_cg.
:pr:16266 by :user:Rushabh Vasani <rushabh-v>.
|API| Deprecated public attributes standard_coef_, standard_intercept_,
average_coef_, and average_intercept_ in
:class:linear_model.SGDClassifier,
:class:linear_model.SGDRegressor,
:class:linear_model.PassiveAggressiveClassifier,
:class:linear_model.PassiveAggressiveRegressor.
:pr:16261 by :user:Carlos Brandt <chbrandt>.
|Fix| |Efficiency| :class:linear_model.ARDRegression is more stable and
much faster when n_samples > n_features. It can now scale to hundreds of
thousands of samples. The stability fix might imply changes in the number
of non-zero coefficients and in the predicted output. :pr:16849 by
Nicolas Hug_.
|Fix| Fixed a bug in :class:linear_model.ElasticNetCV,
:class:linear_model.MultiTaskElasticNetCV, :class:linear_model.LassoCV
and :class:linear_model.MultiTaskLassoCV where fitting would fail when
using joblib loky backend. :pr:14264 by
:user:Jérémie du Boisberranger <jeremiedbb>.
|Efficiency| Speed up :class:linear_model.MultiTaskLasso,
:class:linear_model.MultiTaskLassoCV, :class:linear_model.MultiTaskElasticNet,
:class:linear_model.MultiTaskElasticNetCV by avoiding slower
BLAS Level 2 calls on small arrays
:pr:17021 by :user:Alex Gramfort <agramfort> and
:user:Mathurin Massias <mathurinm>.
:mod:sklearn.metrics
......................
|Enhancement| :func:metrics.pairwise_distances_chunked now allows
its reduce_func to not have a return value, enabling in-place operations.
:pr:16397 by Joel Nothman_.
|Fix| Fixed a bug in :func:metrics.mean_squared_error to not ignore
argument squared when argument multioutput='raw_values'.
:pr:16323 by :user:Rushabh Vasani <rushabh-v>
|Fix| Fixed a bug in :func:metrics.mutual_info_score where negative
scores could be returned. :pr:16362 by Thomas Fan_.
|Fix| Fixed a bug in :func:metrics.confusion_matrix that would raise
an error when y_true and y_pred were length zero and labels was
not None. In addition, we raise an error when an empty list is given to
the labels parameter.
:pr:16442 by :user:Kyle Parsons <parsons-kyle-89>.
|API| Changed the formatting of values in
:meth:metrics.ConfusionMatrixDisplay.plot and
metrics.plot_confusion_matrix to pick the shorter format (either '2g'
or 'd'). :pr:16159 by :user:Rick Mackenbach <Rick-Mackenbach> and
Thomas Fan_.
|API| From version 0.25, :func:metrics.pairwise_distances will no
longer automatically compute the VI parameter for Mahalanobis distance
and the V parameter for seuclidean distance if Y is passed. The user
will be expected to compute this parameter on the training data of their
choice and pass it to pairwise_distances. :pr:16993 by Joel Nothman_.
:mod:sklearn.model_selection
..............................
|Enhancement| :class:model_selection.GridSearchCV and
:class:model_selection.RandomizedSearchCV yields stack trace information
in fit failed warning messages in addition to previously emitted
type and details.
:pr:15622 by :user:Gregory Morse <GregoryMorse>.
|Fix| :func:model_selection.cross_val_predict supports
method="predict_proba" when y=None. :pr:15918 by
:user:Luca Kubin <lkubin>.
|Fix| model_selection.fit_grid_point is deprecated in 0.23 and will
be removed in 0.25. :pr:16401 by
:user:Arie Pratama Sutiono <ariepratama>
:mod:sklearn.multioutput
..........................
|Feature| :func:multioutput.MultiOutputRegressor.fit and
:func:multioutput.MultiOutputClassifier.fit now can accept fit_params
to pass to the estimator.fit method of each step. :issue:15953
:pr:15959 by :user:Ke Huang <huangk10>.
|Enhancement| :class:multioutput.RegressorChain now supports fit_params
for base_estimator during fit.
:pr:16111 by :user:Venkatachalam N <venkyyuvy>.
:mod:sklearn.naive_bayes
.............................
naive_bayes.CategoricalNB when the number of features in the input
differs between predict and fit.
:pr:16090 by :user:Madhura Jayaratne <madhuracj>.:mod:sklearn.neural_network
.............................
|Efficiency| :class:neural_network.MLPClassifier and
:class:neural_network.MLPRegressor has reduced memory footprint when using
stochastic solvers, 'sgd' or 'adam', and shuffle=True. :pr:14075 by
:user:meyer89.
|Fix| Increases the numerical stability of the logistic loss function in
:class:neural_network.MLPClassifier by clipping the probabilities.
:pr:16117 by Thomas Fan_.
:mod:sklearn.inspection
.........................
inspection.PartialDependenceDisplay now exposes the
deciles lines as attributes so they can be hidden or customized. :pr:15785
by Nicolas Hug_:mod:sklearn.preprocessing
............................
|Feature| argument drop of :class:preprocessing.OneHotEncoder
will now accept value 'if_binary' and will drop the first category of
each feature with two categories. :pr:16245
by :user:Rushabh Vasani <rushabh-v>.
|Enhancement| :class:preprocessing.OneHotEncoder's drop_idx_ ndarray
can now contain None, where drop_idx_[i] = None means that no category
is dropped for index i. :pr:16585 by :user:Chiara Marmo <cmarmo>.
|Enhancement| :class:preprocessing.MaxAbsScaler,
:class:preprocessing.MinMaxScaler, :class:preprocessing.StandardScaler,
:class:preprocessing.PowerTransformer,
:class:preprocessing.QuantileTransformer,
:class:preprocessing.RobustScaler now supports pandas' nullable integer
dtype with missing values. :pr:16508 by Thomas Fan_.
|Efficiency| :class:preprocessing.OneHotEncoder is now faster at
transforming. :pr:15762 by Thomas Fan_.
|Fix| Fix a bug in :class:preprocessing.StandardScaler which was incorrectly
computing statistics when calling partial_fit on sparse inputs.
:pr:16466 by :user:Guillaume Lemaitre <glemaitre>.
|Fix| Fix a bug in :class:preprocessing.Normalizer with norm='max',
which was not taking the absolute value of the maximum values before
normalizing the vectors. :pr:16632 by
:user:Maura Pintor <Maupin1991> and :user:Battista Biggio <bbiggio>.
:mod:sklearn.semi_supervised
..............................
semi_supervised.LabelSpreading and
:class:semi_supervised.LabelPropagation avoids divide by zero warnings
when normalizing label_distributions_. :pr:15946 by :user:ngshya.:mod:sklearn.svm
..................
|Fix| |Efficiency| Improved libsvm and liblinear random number
generators used to randomly select coordinates in the coordinate descent
algorithms. Platform-dependent C rand() was used, which is only able to
generate numbers up to 32767 on windows platform (see this blog post <https://codeforces.com/blog/entry/61587>) and also has poor
randomization power as suggested by this presentation <https://channel9.msdn.com/Events/GoingNative/2013/rand-Considered-Harmful>.
It was replaced with C++11 mt19937, a Mersenne Twister that correctly
generates 31bits/63bits random numbers on all platforms. In addition, the
crude "modulo" postprocessor used to get a random number in a bounded
interval was replaced by the tweaked Lemire method as suggested by this blog post <https://www.pcg-random.org/posts/bounded-rands.html>_.
Any model using the svm.libsvm or the svm.liblinear solver,
including :class:svm.LinearSVC, :class:svm.LinearSVR,
:class:svm.NuSVC, :class:svm.NuSVR, :class:svm.OneClassSVM,
:class:svm.SVC, :class:svm.SVR, :class:linear_model.LogisticRegression,
is affected. In particular users can expect a better convergence when the
number of samples (LibSVM) or the number of features (LibLinear) is large.
:pr:13511 by :user:Sylvain Marié <smarie>.
|Fix| Fix use of custom kernel not taking float entries such as string
kernels in :class:svm.SVC and :class:svm.SVR. Note that custom kernels
are now expected to validate their input where they previously received
valid numeric arrays.
:pr:11296 by Alexandre Gramfort_ and :user:Georgi Peev <georgipeev>.
|API| :class:svm.SVR and :class:svm.OneClassSVM attributes, probA_ and
probB_, are now deprecated as they were not useful. :pr:15558 by
Thomas Fan_.
:mod:sklearn.tree
...................
|Fix| :func:tree.plot_tree rotate parameter was unused and has been
deprecated.
:pr:15806 by :user:Chiara Marmo <cmarmo>.
|Fix| Fix support of read-only float32 array input in predict,
decision_path and predict_proba methods of
:class:tree.DecisionTreeClassifier, :class:tree.ExtraTreeClassifier and
:class:ensemble.GradientBoostingClassifier as well as predict method of
:class:tree.DecisionTreeRegressor, :class:tree.ExtraTreeRegressor, and
:class:ensemble.GradientBoostingRegressor.
:pr:16331 by :user:Alexandre Batisse <batalex>.
:mod:sklearn.utils
....................
|MajorFeature| Estimators can now be displayed with a rich html
representation. This can be enabled in Jupyter notebooks by setting
display='diagram' in :func:~sklearn.set_config. The raw html can be
returned by using :func:utils.estimator_html_repr.
:pr:14180 by Thomas Fan_.
|Enhancement| improve error message in :func:utils.validation.column_or_1d.
:pr:15926 by :user:Loïc Estève <lesteve>.
|Enhancement| add warning in :func:utils.check_array for
pandas sparse DataFrame.
:pr:16021 by :user:Rushabh Vasani <rushabh-v>.
|Enhancement| :func:utils.check_array now constructs a sparse
matrix from a pandas DataFrame that contains only SparseArray columns.
:pr:16728 by Thomas Fan_.
|Enhancement| :func:utils.check_array supports pandas'
nullable integer dtype with missing values when force_all_finite is set to
False or 'allow-nan' in which case the data is converted to floating
point values where pd.NA values are replaced by np.nan. As a consequence,
all :mod:sklearn.preprocessing transformers that accept numeric inputs with
missing values represented as np.nan now also accepts being directly fed
pandas dataframes with pd.Int* or pd.Uint*typed columns that usepd.NA as a missing value marker. :pr:16508byThomas Fan`_.
|API| Passing classes to :func:utils.estimator_checks.check_estimator and
:func:utils.estimator_checks.parametrize_with_checks is now deprecated,
and support for classes will be removed in 0.24. Pass instances instead.
:pr:17032 by Nicolas Hug_.
|API| The private utility _safe_tags in utils.estimator_checks was
removed, hence all tags should be obtained through estimator._get_tags().
Note that Mixins like RegressorMixin must come before base classes
in the MRO for _get_tags() to work properly.
:pr:16950 by Nicolas Hug_.
|FIX| utils.all_estimators now only returns public estimators.
:pr:15380 by Thomas Fan_.
Miscellaneous .............
|MajorFeature| Adds a HTML representation of estimators to be shown in
a jupyter notebook or lab. This visualization is activated by setting the
display option in :func:sklearn.set_config. :pr:14180 by
Thomas Fan_.
|Enhancement| scikit-learn now works with mypy without errors.
:pr:16726 by Roman Yurchak_.
|API| Most estimators now expose a n_features_in_ attribute. This
attribute is equal to the number of features passed to the fit method.
See SLEP010 <https://scikit-learn-enhancement-proposals.readthedocs.io/en/latest/slep010/proposal.html>_
for details. :pr:16112 by Nicolas Hug_.
|API| Estimators now have a requires_y tags which is False by default
except for estimators that inherit from ~sklearn.base.RegressorMixin or
~sklearn.base.ClassifierMixin. This tag is used to ensure that a proper
error message is raised when y was expected but None was passed.
:pr:16622 by Nicolas Hug_.
|API| The default setting print_changed_only has been changed from False
to True. This means that the repr of estimators is now more concise and
only shows the parameters whose default value has been changed when
printing an estimator. You can restore the previous behaviour by using
sklearn.set_config(print_changed_only=False). Also, note that it is
always possible to quickly inspect the parameters of any estimator using
est.get_params(deep=False). :pr:17061 by Nicolas Hug_.
.. rubric:: Code and documentation contributors
Thanks to everyone who has contributed to the maintenance and improvement of the project since version 0.22, including:
Abbie Popa, Adrin Jalali, Aleksandra Kocot, Alexandre Batisse, Alexandre Gramfort, Alex Henrie, Alex Itkes, Alex Liang, alexshacked, Alonso Silva Allende, Ana Casado, Andreas Mueller, Angela Ambroz, Ankit810, Arie Pratama Sutiono, Arunav Konwar, Baptiste Maingret, Benjamin Beier Liu, bernie gray, Bharathi Srinivasan, Bharat Raghunathan, Bibhash Chandra Mitra, Brian Wignall, brigi, Brigitta Sipőcz, Carlos H Brandt, CastaChick, castor, cgsavard, Chiara Marmo, Chris Gregory, Christian Kastner, Christian Lorentzen, Corrie Bartelheimer, Daniël van Gelder, Daphne, David Breuer, david-cortes, dbauer9, Divyaprabha M, Edward Qian, Ekaterina Borovikova, ELNS, Emily Taylor, Erich Schubert, Eric Leung, Evgeni Chasnovski, Fabiana, Facundo Ferrín, Fan, Franziska Boenisch, Gael Varoquaux, Gaurav Sharma, Geoffrey Bolmier, Georgi Peev, gholdman1, Gonthier Nicolas, Gregory Morse, Gregory R. Lee, Guillaume Lemaitre, Gui Miotto, Hailey Nguyen, Hanmin Qin, Hao Chun Chang, HaoYin, Hélion du Mas des Bourboux, Himanshu Garg, Hirofumi Suzuki, huangk10, Hugo van Kemenade, Hye Sung Jung, indecisiveuser, inderjeet, J-A16, Jérémie du Boisberranger, Jin-Hwan CHO, JJmistry, Joel Nothman, Johann Faouzi, Jon Haitz Legarreta Gorroño, Juan Carlos Alfaro Jiménez, judithabk6, jumon, Kathryn Poole, Katrina Ni, Kesshi Jordan, Kevin Loftis, Kevin Markham, krishnachaitanya9, Lam Gia Thuan, Leland McInnes, Lisa Schwetlick, lkubin, Loic Esteve, lopusz, lrjball, lucgiffon, lucyleeow, Lucy Liu, Lukas Kemkes, Maciej J Mikulski, Madhura Jayaratne, Magda Zielinska, maikia, Mandy Gu, Manimaran, Manish Aradwad, Maren Westermann, Maria, Mariana Meireles, Marie Douriez, Marielle, Mateusz Górski, mathurinm, Matt Hall, Maura Pintor, mc4229, meyer89, m.fab, Michael Shoemaker, Michał Słapek, Mina Naghshhnejad, mo, Mohamed Maskani, Mojca Bertoncelj, narendramukherjee, ngshya, Nicholas Won, Nicolas Hug, nicolasservel, Niklas, @nkish, Noa Tamir, Oleksandr Pavlyk, olicairns, Oliver Urs Lenz, Olivier Grisel, parsons-kyle-89, Paula, Pete Green, Pierre Delanoue, pspachtholz, Pulkit Mehta, Qizhi Jiang, Quang Nguyen, rachelcjordan, raduspaimoc, Reshama Shaikh, Riccardo Folloni, Rick Mackenbach, Ritchie Ng, Roman Feldbauer, Roman Yurchak, Rory Hartong-Redden, Rüdiger Busche, Rushabh Vasani, Sambhav Kothari, Samesh Lakhotia, Samuel Duan, SanthoshBala18, Santiago M. Mola, Sarat Addepalli, scibol, Sebastian Kießling, SergioDSR, Sergul Aydore, Shiki-H, shivamgargsya, SHUBH CHATTERJEE, Siddharth Gupta, simonamaggio, smarie, Snowhite, stareh, Stephen Blystone, Stephen Marsh, Sunmi Yoon, SylvainLan, talgatomarov, tamirlan1, th0rwas, theoptips, Thomas J Fan, Thomas Li, Thomas Schmitt, Tim Nonner, Tim Vink, Tiphaine Viard, Tirth Patel, Titus Christian, Tom Dupré la Tour, trimeta, Vachan D A, Vandana Iyer, Venkatachalam N, waelbenamara, wconnell, wderose, wenliwyan, Windber, wornbb, Yu-Hang "Maxin" Tang