# Version 0.21.3¶

## Legend for changelogs¶

Major Feature : something big that you couldn’t do before.

Feature : something that you couldn’t do before.

Efficiency : an existing feature now may not require as much computation or memory.

Enhancement : a miscellaneous minor improvement.

Fix : something that previously didn’t work as documentated – or according to reasonable expectations – should now work.

API Change : you will need to change your code to have the same effect in the future; or a feature will be removed in the future.

**July 30, 2019**

## Changed models¶

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.

The v0.20.0 release notes failed to mention a backwards incompatibility in

`metrics.make_scorer`

when`needs_proba=True`

and`y_true`

is binary. Now, the scorer function is supposed to accept a 1D`y_pred`

(i.e., probability of the positive class, shape`(n_samples,)`

), instead of a 2D`y_pred`

(i.e., shape`(n_samples, 2)`

).

## Changelog¶

`sklearn.cluster`

¶

Fix Fixed a bug in

`cluster.KMeans`

where computation with`init='random'`

was single threaded for`n_jobs > 1`

or`n_jobs = -1`

. #12955 by Prabakaran Kumaresshan.Fix Fixed a bug in

`cluster.OPTICS`

where users were unable to pass float`min_samples`

and`min_cluster_size`

. #14496 by Fabian Klopfer and Hanmin Qin.Fix Fixed a bug in

`cluster.KMeans`

where KMeans++ initialisation could rarely result in an IndexError. #11756 by Joel Nothman.

`sklearn.compose`

¶

Fix Fixed an issue in

`compose.ColumnTransformer`

where using DataFrames whose column order differs between :func:`fit`

and :func:`transform`

could lead to silently passing incorrect columns to the`remainder`

transformer. #14237 by`Andreas Schuderer <schuderer>`

.

`sklearn.datasets`

¶

Fix

`datasets.fetch_california_housing`

,`datasets.fetch_covtype`

,`datasets.fetch_kddcup99`

,`datasets.fetch_olivetti_faces`

,`datasets.fetch_rcv1`

, and`datasets.fetch_species_distributions`

try to persist the previously cache using the new`joblib`

if the cached data was persisted using the deprecated`sklearn.externals.joblib`

. This behavior is set to be deprecated and removed in v0.23. #14197 by Adrin Jalali.

`sklearn.ensemble`

¶

Fix Fix zero division error in

`HistGradientBoostingClassifier`

and`HistGradientBoostingRegressor`

. #14024 by`Nicolas Hug <NicolasHug>`

.

`sklearn.impute`

¶

Fix Fixed a bug in

`impute.SimpleImputer`

and`impute.IterativeImputer`

so that no errors are thrown when there are missing values in training data. #13974 by`Frank Hoang <fhoang7>`

.

`sklearn.inspection`

¶

Fix Fixed a bug in

`inspection.plot_partial_dependence`

where`target`

parameter was not being taken into account for multiclass problems. #14393 by Guillem G. Subies.

`sklearn.linear_model`

¶

Fix Fixed a bug in

`linear_model.LogisticRegressionCV`

where`refit=False`

would fail depending on the`'multiclass'`

and`'penalty'`

parameters (regression introduced in 0.21). #14087 by Nicolas Hug.Fix Compatibility fix for

`linear_model.ARDRegression`

and Scipy>=1.3.0. Adapts to upstream changes to the default`pinvh`

cutoff threshold which otherwise results in poor accuracy in some cases. #14067 by Tim Staley.

`sklearn.neighbors`

¶

Fix Fixed a bug in

`neighbors.NeighborhoodComponentsAnalysis`

where the validation of initial parameters`n_components`

,`max_iter`

and`tol`

required too strict types. #14092 by Jérémie du Boisberranger.

`sklearn.tree`

¶

Fix Fixed bug in

`tree.export_text`

when the tree has one feature and a single feature name is passed in. #14053 by`Thomas Fan`

.Fix Fixed an issue with

`plot_tree`

where it displayed entropy calculations even for`gini`

criterion in DecisionTreeClassifiers. #13947 by Frank Hoang.

# Version 0.21.2¶

**24 May 2019**

## Changelog¶

`sklearn.decomposition`

¶

Fix Fixed a bug in

`cross_decomposition.CCA`

improving numerical stability when`Y`

is close to zero. #13903 by Thomas Fan.

`sklearn.metrics`

¶

Fix Fixed a bug in

`metrics.pairwise.euclidean_distances`

where a part of the distance matrix was left un-instanciated for suffiently large float32 datasets (regression introduced in 0.21). #13910 by Jérémie du Boisberranger.

`sklearn.preprocessing`

¶

Fix Fixed a bug in

`preprocessing.OneHotEncoder`

where the new`drop`

parameter was not reflected in`get_feature_names`

. #13894 by James Myatt.

`sklearn.utils.sparsefuncs`

¶

Fix Fixed a bug where

`min_max_axis`

would fail on 32-bit systems for certain large inputs. This affects`preprocessing.MaxAbsScaler`

,`preprocessing.normalize`

and`preprocessing.LabelBinarizer`

. #13741 by Roddy MacSween.

# Version 0.21.1¶

**17 May 2019**

This is a bug-fix release to primarily resolve some packaging issues in version 0.21.0. It also includes minor documentation improvements and some bug fixes.

## Changelog¶

`sklearn.inspection`

¶

Fix Fixed a bug in

`inspection.partial_dependence`

to only check classifier and not regressor for the multiclass-multioutput case. #14309 by Guillaume Lemaitre.

`sklearn.metrics`

¶

Fix Fixed a bug in

`metrics.pairwise_distances`

where it would raise`AttributeError`

for boolean metrics when`X`

had a boolean dtype and`Y == None`

. #13864 by Paresh Mathur.Fix Fixed two bugs in

`metrics.pairwise_distances`

when`n_jobs > 1`

. First it used to return a distance matrix with same dtype as input, even for integer dtype. Then the diagonal was not zeros for euclidean metric when`Y`

is`X`

. #13877 by Jérémie du Boisberranger.

`sklearn.neighbors`

¶

Fix Fixed a bug in

`neighbors.KernelDensity`

which could not be restored from a pickle if`sample_weight`

had been used. #13772 by Aditya Vyas.

# Version 0.21.0¶

**May 2019**

## Changed models¶

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.

`discriminant_analysis.LinearDiscriminantAnalysis`

for multiclass classification. Fix`discriminant_analysis.LinearDiscriminantAnalysis`

with ‘eigen’ solver. FixDecision trees and derived ensembles when both

`max_depth`

and`max_leaf_nodes`

are set. Fix`linear_model.LogisticRegression`

and`linear_model.LogisticRegressionCV`

with ‘saga’ solver. Fix`sklearn.feature_extraction.text.HashingVectorizer`

,`sklearn.feature_extraction.text.TfidfVectorizer`

, and`sklearn.feature_extraction.text.CountVectorizer`

Fix`svm.SVC.decision_function`

and`multiclass.OneVsOneClassifier.decision_function`

. Fix`linear_model.SGDClassifier`

and any derived classifiers. FixAny model using the

`linear_model._sag.sag_solver`

function with a`0`

seed, including`linear_model.LogisticRegression`

,`linear_model.LogisticRegressionCV`

,`linear_model.Ridge`

, and`linear_model.RidgeCV`

with ‘sag’ solver. Fix`linear_model.RidgeCV`

when using generalized cross-validation with sparse inputs. Fix

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.)

## Known Major Bugs¶

The default

`max_iter`

for`linear_model.LogisticRegression`

is too small for many solvers given the default`tol`

. In particular, we accidentally changed the default`max_iter`

for the liblinear solver from 1000 to 100 iterations in #3591 released in version 0.16. In a future release we hope to choose better default`max_iter`

and`tol`

heuristically depending on the solver (see #13317).

## Changelog¶

Support for Python 3.4 and below has been officially dropped.

`sklearn.base`

¶

API Change The R2 score used when calling

`score`

on a regressor will use`multioutput='uniform_average'`

from version 0.23 to keep consistent with`metrics.r2_score`

. This will influence the`score`

method of all the multioutput regressors (except for`multioutput.MultiOutputRegressor`

). #13157 by Hanmin Qin.

`sklearn.calibration`

¶

Enhancement Added support to bin the data passed into

`calibration.calibration_curve`

by quantiles instead of uniformly between 0 and 1. #13086 by Scott Cole.Enhancement Allow n-dimensional arrays as input for

`calibration.CalibratedClassifierCV`

. #13485 by William de Vazelhes.

`sklearn.cluster`

¶

Major Feature A new clustering algorithm:

`cluster.OPTICS`

: an algorithm related to`cluster.DBSCAN`

, that has hyperparameters easier to set and that scales better, by Shane, Adrin Jalali, Erich Schubert, Hanmin Qin, and Assia Benbihi.Fix Fixed a bug where

`cluster.Birch`

could occasionally raise an AttributeError. #13651 by Joel Nothman.Fix Fixed a bug in

`cluster.KMeans`

where empty clusters weren’t correctly relocated when using sample weights. #13486 by Jérémie du Boisberranger.API Change The

`n_components_`

attribute in`cluster.AgglomerativeClustering`

and`cluster.FeatureAgglomeration`

has been renamed to`n_connected_components_`

. #13427 by Stephane Couvreur.Enhancement

`cluster.AgglomerativeClustering`

and`cluster.FeatureAgglomeration`

now accept a`distance_threshold`

parameter which can be used to find the clusters instead of`n_clusters`

. #9069 by Vathsala Achar and Adrin Jalali.

`sklearn.compose`

¶

API Change

`compose.ColumnTransformer`

is no longer an experimental feature. #13835 by Hanmin Qin.

`sklearn.datasets`

¶

Fix Added support for 64-bit group IDs and pointers in SVMLight files. #10727 by Bryan K Woods.

Fix

`datasets.load_sample_images`

returns images with a deterministic order. #13250 by Thomas Fan.

`sklearn.decomposition`

¶

Enhancement

`decomposition.KernelPCA`

now has deterministic output (resolved sign ambiguity in eigenvalue decomposition of the kernel matrix). #13241 by Aurélien Bellet.Fix Fixed a bug in

`decomposition.KernelPCA`

,`fit().transform()`

now produces the correct output (the same as`fit_transform()`

) in case of non-removed zero eigenvalues (`remove_zero_eig=False`

).`fit_inverse_transform`

was also accelerated by using the same trick as`fit_transform`

to compute the transform of`X`

. #12143 by Sylvain MariéFix Fixed a bug in

`decomposition.NMF`

where`init = 'nndsvd'`

,`init = 'nndsvda'`

, and`init = 'nndsvdar'`

are allowed when`n_components < n_features`

instead of`n_components <= min(n_samples, n_features)`

. #11650 by Hossein Pourbozorg and Zijie (ZJ) Poh.API Change The default value of the

`init`

argument in`decomposition.non_negative_factorization`

will change from`random`

to`None`

in version 0.23 to make it consistent with`decomposition.NMF`

. A FutureWarning is raised when the default value is used. #12988 by Zijie (ZJ) Poh.

`sklearn.discriminant_analysis`

¶

Enhancement

`discriminant_analysis.LinearDiscriminantAnalysis`

now preserves`float32`

and`float64`

dtypes. #8769 and #11000 by Thibault SejourneFix A

`ChangedBehaviourWarning`

is now raised when`discriminant_analysis.LinearDiscriminantAnalysis`

is given as parameter`n_components > min(n_features, n_classes - 1)`

, and`n_components`

is changed to`min(n_features, n_classes - 1)`

if so. Previously the change was made, but silently. #11526 by William de Vazelhes.Fix Fixed a bug in

`discriminant_analysis.LinearDiscriminantAnalysis`

where the predicted probabilities would be incorrectly computed in the multiclass case. #6848, by Agamemnon Krasoulis and`Guillaume Lemaitre <glemaitre>`

.Fix Fixed a bug in

`discriminant_analysis.LinearDiscriminantAnalysis`

where the predicted probabilities would be incorrectly computed with`eigen`

solver. #11727, by Agamemnon Krasoulis.

`sklearn.dummy`

¶

Fix Fixed a bug in

`dummy.DummyClassifier`

where the`predict_proba`

method was returning int32 array instead of float64 for the`stratified`

strategy. #13266 by Christos Aridas.Fix Fixed a bug in

`dummy.DummyClassifier`

where it was throwing a dimension mismatch error in prediction time if a column vector`y`

with`shape=(n, 1)`

was given at`fit`

time. #13545 by Nick Sorros and Adrin Jalali.

`sklearn.ensemble`

¶

Major Feature Add two new implementations of gradient boosting trees:

`ensemble.HistGradientBoostingClassifier`

and`ensemble.HistGradientBoostingRegressor`

. The implementation of these estimators is inspired by LightGBM and can be orders of magnitude faster than`ensemble.GradientBoostingRegressor`

and`ensemble.GradientBoostingClassifier`

when the number of samples is larger than tens of thousands of samples. The API of these new estimators is slightly different, and some of the features from`ensemble.GradientBoostingClassifier`

and`ensemble.GradientBoostingRegressor`

are not yet supported.These new estimators are experimental, which means that their results or their API might change without any deprecation cycle. To use them, you need to explicitly import

`enable_hist_gradient_boosting`

:>>> # explicitly require this experimental feature >>> from sklearn.experimental import enable_hist_gradient_boosting # noqa >>> # now you can import normally from sklearn.ensemble >>> from sklearn.ensemble import HistGradientBoostingClassifier

#12807 by Nicolas Hug.

Feature Add

`ensemble.VotingRegressor`

which provides an equivalent of`ensemble.VotingClassifier`

for regression problems. #12513 by Ramil Nugmanov and Mohamed Ali Jamaoui.Efficiency Make

`ensemble.IsolationForest`

prefer threads over processes when running with`n_jobs > 1`

as the underlying decision tree fit calls do release the GIL. This changes reduces memory usage and communication overhead. #12543 by Isaac Storch and Olivier Grisel.Efficiency Make

`ensemble.IsolationForest`

more memory efficient by avoiding keeping in memory each tree prediction. #13260 by Nicolas Goix.Efficiency

`ensemble.IsolationForest`

now uses chunks of data at prediction step, thus capping the memory usage. #13283 by Nicolas Goix.Efficiency

`sklearn.ensemble.GradientBoostingClassifier`

and`sklearn.ensemble.GradientBoostingRegressor`

now keep the input`y`

as`float64`

to avoid it being copied internally by trees. #13524 by Adrin Jalali.Enhancement Minimized the validation of X in

`ensemble.AdaBoostClassifier`

and`ensemble.AdaBoostRegressor`

#13174 by Christos Aridas.Enhancement

`ensemble.IsolationForest`

now exposes`warm_start`

parameter, allowing iterative addition of trees to an isolation forest. #13496 by Peter Marko.Fix The values of

`feature_importances_`

in all random forest based models (i.e.`ensemble.RandomForestClassifier`

,`ensemble.RandomForestRegressor`

,`ensemble.ExtraTreesClassifier`

,`ensemble.ExtraTreesRegressor`

,`ensemble.RandomTreesEmbedding`

,`ensemble.GradientBoostingClassifier`

, and`ensemble.GradientBoostingRegressor`

) now:sum up to

`1`

all the single node trees in feature importance calculation are ignored

in case all trees have only one single node (i.e. a root node), feature importances will be an array of all zeros.

#13636 and #13620 by Adrin Jalali.

Fix Fixed a bug in

`ensemble.GradientBoostingClassifier`

and`ensemble.GradientBoostingRegressor`

, which didn’t support scikit-learn estimators as the initial estimator. Also added support of initial estimator which does not support sample weights. #12436 by Jérémie du Boisberranger and #12983 by Nicolas Hug.Fix Fixed the output of the average path length computed in

`ensemble.IsolationForest`

when the input is either 0, 1 or 2. #13251 by Albert Thomas and joshuakennethjones.Fix Fixed a bug in

`ensemble.GradientBoostingClassifier`

where the gradients would be incorrectly computed in multiclass classification problems. #12715 by Nicolas Hug.Fix Fixed a bug in

`ensemble.GradientBoostingClassifier`

where validation sets for early stopping were not sampled with stratification. #13164 by Nicolas Hug.Fix Fixed a bug in

`ensemble.GradientBoostingClassifier`

where the default initial prediction of a multiclass classifier would predict the classes priors instead of the log of the priors. #12983 by Nicolas Hug.Fix Fixed a bug in

`ensemble.RandomForestClassifier`

where the`predict`

method would error for multiclass multioutput forests models if any targets were strings. #12834 by Elizabeth Sander.Fix Fixed a bug in

`ensemble.gradient_boosting.LossFunction`

and`ensemble.gradient_boosting.LeastSquaresError`

where the default value of`learning_rate`

in`update_terminal_regions`

is not consistent with the document and the caller functions. Note however that directly using these loss functions is deprecated. #6463 by movelikeriver.Fix

`ensemble.partial_dependence`

(and consequently the new version`sklearn.inspection.partial_dependence`

) now takes sample weights into account for the partial dependence computation when the gradient boosting model has been trained with sample weights. #13193 by Samuel O. Ronsin.API Change

`ensemble.partial_dependence`

and`ensemble.plot_partial_dependence`

are now deprecated in favor of`inspection.partial_dependence`

and`inspection.plot_partial_dependence`

. #12599 by Trevor Stephens and Nicolas Hug.Fix

`ensemble.VotingClassifier`

and`ensemble.VotingRegressor`

were failing during`fit`

in one of the estimators was set to`None`

and`sample_weight`

was not`None`

. #13779 by Guillaume Lemaitre.API Change

`ensemble.VotingClassifier`

and`ensemble.VotingRegressor`

accept`'drop'`

to disable an estimator in addition to`None`

to be consistent with other estimators (i.e.,`pipeline.FeatureUnion`

and`compose.ColumnTransformer`

). #13780 by Guillaume Lemaitre.

`sklearn.externals`

¶

API Change Deprecated

`externals.six`

since we have dropped support for Python 2.7. #12916 by Hanmin Qin.

`sklearn.feature_extraction`

¶

Fix If

`input='file'`

or`input='filename'`

, and a callable is given as the`analyzer`

,`sklearn.feature_extraction.text.HashingVectorizer`

,`sklearn.feature_extraction.text.TfidfVectorizer`

, and`sklearn.feature_extraction.text.CountVectorizer`

now read the data from the file(s) and then pass it to the given`analyzer`

, instead of passing the file name(s) or the file object(s) to the analyzer. #13641 by Adrin Jalali.

`sklearn.impute`

¶

Major Feature Added

`impute.IterativeImputer`

, which is a strategy for imputing missing values by modeling each feature with missing values as a function of other features in a round-robin fashion. #8478 and #12177 by Sergey Feldman and Ben Lawson.The API of IterativeImputer is experimental and subject to change without any deprecation cycle. To use them, you need to explicitly import

`enable_iterative_imputer`

:>>> from sklearn.experimental import enable_iterative_imputer # noqa >>> # now you can import normally from sklearn.impute >>> from sklearn.impute import IterativeImputer

Feature The

`impute.SimpleImputer`

and`impute.IterativeImputer`

have a new parameter`'add_indicator'`

, which simply stacks a`impute.MissingIndicator`

transform into the output of the imputer’s transform. That allows a predictive estimator to account for missingness. #12583, #13601 by Danylo Baibak.Fix In

`impute.MissingIndicator`

avoid implicit densification by raising an exception if input is sparse add`missing_values`

property is set to 0. #13240 by Bartosz Telenczuk.Fix Fixed two bugs in

`impute.MissingIndicator`

. First, when`X`

is sparse, all the non-zero non missing values used to become explicit False in the transformed data. Then, when`features='missing-only'`

, all features used to be kept if there were no missing values at all. #13562 by Jérémie du Boisberranger.

`sklearn.inspection`

¶

(new subpackage)

Feature Partial dependence plots (

`inspection.plot_partial_dependence`

) are now supported for any regressor or classifier (provided that they have a`predict_proba`

method). #12599 by Trevor Stephens and Nicolas Hug.

`sklearn.isotonic`

¶

Feature Allow different dtypes (such as float32) in

`isotonic.IsotonicRegression`

. #8769 by Vlad Niculae

`sklearn.linear_model`

¶

Enhancement

`linear_model.Ridge`

now preserves`float32`

and`float64`

dtypes. #8769 and #11000 by Guillaume Lemaitre, and Joan MassichFeature

`linear_model.LogisticRegression`

and`linear_model.LogisticRegressionCV`

now support Elastic-Net penalty, with the ‘saga’ solver. #11646 by Nicolas Hug.Feature Added

`linear_model.lars_path_gram`

, which is`linear_model.lars_path`

in the sufficient stats mode, allowing users to compute`linear_model.lars_path`

without providing`X`

and`y`

. #11699 by Kuai Yu.Efficiency

`linear_model.make_dataset`

now preserves`float32`

and`float64`

dtypes, reducing memory consumption in stochastic gradient, SAG and SAGA solvers. #8769 and #11000 by Nelle Varoquaux, Arthur Imbert, Guillaume Lemaitre, and Joan MassichEnhancement

`linear_model.LogisticRegression`

now supports an unregularized objective when`penalty='none'`

is passed. This is equivalent to setting`C=np.inf`

with l2 regularization. Not supported by the liblinear solver. #12860 by Nicolas Hug.Enhancement

`sparse_cg`

solver in`linear_model.Ridge`

now supports fitting the intercept (i.e.`fit_intercept=True`

) when inputs are sparse. #13336 by Bartosz Telenczuk.Enhancement The coordinate descent solver used in

`Lasso`

,`ElasticNet`

, etc. now issues a`ConvergenceWarning`

when it completes without meeting the desired toleranbce. #11754 and #13397 by Brent Fagan and Adrin Jalali.Fix Fixed a bug in

`linear_model.LogisticRegression`

and`linear_model.LogisticRegressionCV`

with ‘saga’ solver, where the weights would not be correctly updated in some cases. #11646 by Tom Dupre la Tour.Fix Fixed the posterior mean, posterior covariance and returned regularization parameters in

`linear_model.BayesianRidge`

. The posterior mean and the posterior covariance were not the ones computed with the last update of the regularization parameters and the returned regularization parameters were not the final ones. Also fixed the formula of the log marginal likelihood used to compute the score when`compute_score=True`

. #12174 by Albert Thomas.Fix Fixed a bug in

`linear_model.LassoLarsIC`

, where user input`copy_X=False`

at instance creation would be overridden by default parameter value`copy_X=True`

in`fit`

. #12972 by Lucio Fernandez-ArjonaFix Fixed a bug in

`linear_model.LinearRegression`

that was not returning the same coeffecients and intercepts with`fit_intercept=True`

in sparse and dense case. #13279 by Alexandre GramfortFix Fixed a bug in

`linear_model.HuberRegressor`

that was broken when`X`

was of dtype bool. #13328 by Alexandre Gramfort.Fix Fixed a performance issue of

`saga`

and`sag`

solvers when called in a`joblib.Parallel`

setting with`n_jobs > 1`

and`backend="threading"`

, causing them to perform worse than in the sequential case. #13389 by Pierre Glaser.Fix Fixed a bug in

`linear_model.stochastic_gradient.BaseSGDClassifier`

that was not deterministic when trained in a multi-class setting on several threads. #13422 by Clément Doumouro.Fix Fixed bug in

`linear_model.ridge_regression`

,`linear_model.Ridge`

and`linear_model.RidgeClassifier`

that caused unhandled exception for arguments`return_intercept=True`

and`solver=auto`

(default) or any other solver different from`sag`

. #13363 by Bartosz TelenczukFix

`linear_model.ridge_regression`

will now raise an exception if`return_intercept=True`

and solver is different from`sag`

. Previously, only warning was issued. #13363 by Bartosz TelenczukFix

`linear_model.ridge_regression`

will choose`sparse_cg`

solver for sparse inputs when`solver=auto`

and`sample_weight`

is provided (previously`cholesky`

solver was selected). #13363 by Bartosz TelenczukAPI Change The use of

`linear_model.lars_path`

with`X=None`

while passing`Gram`

is deprecated in version 0.21 and will be removed in version 0.23. Use`linear_model.lars_path_gram`

instead. #11699 by Kuai Yu.API Change

`linear_model.logistic_regression_path`

is deprecated in version 0.21 and will be removed in version 0.23. #12821 by Nicolas Hug.Fix

`linear_model.RidgeCV`

with generalized cross-validation now correctly fits an intercept when`fit_intercept=True`

and the design matrix is sparse. #13350 by Jérôme Dockès

`sklearn.manifold`

¶

Efficiency Make

`manifold.tsne.trustworthiness`

use an inverted index instead of an`np.where`

lookup to find the rank of neighbors in the input space. This improves efficiency in particular when computed with lots of neighbors and/or small datasets. #9907 by William de Vazelhes.

`sklearn.metrics`

¶

Feature Added the

`metrics.max_error`

metric and a corresponding`'max_error'`

scorer for single output regression. #12232 by Krishna Sangeeth.Feature Add

`metrics.multilabel_confusion_matrix`

, which calculates a confusion matrix with true positive, false positive, false negative and true negative counts for each class. This facilitates the calculation of set-wise metrics such as recall, specificity, fall out and miss rate. #11179 by Shangwu Yao and Joel Nothman.Feature

`metrics.jaccard_score`

has been added to calculate the Jaccard coefficient as an evaluation metric for binary, multilabel and multiclass tasks, with an interface analogous to`metrics.f1_score`

. #13151 by Gaurav Dhingra and Joel Nothman.Feature Added

`metrics.pairwise.haversine_distances`

which can be accessed with`metric='pairwise'`

through`metrics.pairwise_distances`

and estimators. (Haversine distance was previously available for nearest neighbors calculation.) #12568 by Wei Xue, Emmanuel Arias and Joel Nothman.Efficiency Faster

`metrics.pairwise_distances`

with`n_jobs`

> 1 by using a thread-based backend, instead of process-based backends. #8216 by Pierre Glaser and Romuald MenuetEfficiency The pairwise manhattan distances with sparse input now uses the BLAS shipped with scipy instead of the bundled BLAS. #12732 by Jérémie du Boisberranger

Enhancement Use label

`accuracy`

instead of`micro-average`

on`metrics.classification_report`

to avoid confusion.`micro-average`

is only shown for multi-label or multi-class with a subset of classes because it is otherwise identical to accuracy. #12334 by Emmanuel Arias, Joel Nothman and Andreas MüllerEnhancement Added

`beta`

parameter to`metrics.homogeneity_completeness_v_measure`

and`metrics.v_measure_score`

to configure the tradeoff between homogeneity and completeness. #13607 by Stephane Couvreur and and Ivan Sanchez.Fix The metric

`metrics.r2_score`

is degenerate with a single sample and now it returns NaN and raises`exceptions.UndefinedMetricWarning`

. #12855 by Pawel Sendyk.Fix Fixed a bug where

`metrics.brier_score_loss`

will sometimes return incorrect result when there’s only one class in`y_true`

. #13628 by Hanmin Qin.Fix Fixed a bug in

`metrics.label_ranking_average_precision_score`

where sample_weight wasn’t taken into account for samples with degenerate labels. #13447 by Dan Ellis.API Change The parameter

`labels`

in`metrics.hamming_loss`

is deprecated in version 0.21 and will be removed in version 0.23. #10580 by Reshama Shaikh and Sandra Mitrovic.Fix The function

`metrics.pairwise.euclidean_distances`

, and therefore several estimators with`metric='euclidean'`

, suffered from numerical precision issues with`float32`

features. Precision has been increased at the cost of a small drop of performance. #13554 by @Celelibi and Jérémie du Boisberranger.API Change

`metrics.jaccard_similarity_score`

is deprecated in favour of the more consistent`metrics.jaccard_score`

. The former behavior for binary and multiclass targets is broken. #13151 by Joel Nothman.

`sklearn.mixture`

¶

Fix Fixed a bug in

`mixture.BaseMixture`

and therefore on estimators based on it, i.e.`mixture.GaussianMixture`

and`mixture.BayesianGaussianMixture`

, where`fit_predict`

and`fit.predict`

were not equivalent. #13142 by Jérémie du Boisberranger.

`sklearn.model_selection`

¶

Feature Classes

`GridSearchCV`

and`RandomizedSearchCV`

now allow for refit=callable to add flexibility in identifying the best estimator. See Balance model complexity and cross-validated score. #11354 by Wenhao Zhang, Joel Nothman and Adrin Jalali.Enhancement Classes

`GridSearchCV`

,`RandomizedSearchCV`

, and methods`cross_val_score`

,`cross_val_predict`

,`cross_validate`

, now print train scores when`return_train_scores`

is True and`verbose`

> 2. For`learning_curve`

, and`validation_curve`

only the latter is required. #12613 and #12669 by Marc Torrellas.Enhancement Some CV splitter classes and

`model_selection.train_test_split`

now raise`ValueError`

when the resulting training set is empty. #12861 by Nicolas Hug.Fix Fixed a bug where

`model_selection.StratifiedKFold`

shuffles each class’s samples with the same`random_state`

, making`shuffle=True`

ineffective. #13124 by Hanmin Qin.Fix Added ability for

`model_selection.cross_val_predict`

to handle multi-label (and multioutput-multiclass) targets with`predict_proba`

-type methods. #8773 by Stephen Hoover.Fix Fixed an issue in

`cross_val_predict`

where`method="predict_proba"`

returned always`0.0`

when one of the classes was excluded in a cross-validation fold. #13366 by Guillaume Fournier

`sklearn.multiclass`

¶

Fix Fixed an issue in

`multiclass.OneVsOneClassifier.decision_function`

where the decision_function value of a given sample was different depending on whether the decision_function was evaluated on the sample alone or on a batch containing this same sample due to the scaling used in decision_function. #10440 by Jonathan Ohayon.

`sklearn.multioutput`

¶

Fix Fixed a bug in

`multioutput.MultiOutputClassifier`

where the`predict_proba`

method incorrectly checked for`predict_proba`

attribute in the estimator object. #12222 by Rebekah Kim

`sklearn.neighbors`

¶

Major Feature Added

`neighbors.NeighborhoodComponentsAnalysis`

for metric learning, which implements the Neighborhood Components Analysis algorithm. #10058 by William de Vazelhes and John Chiotellis.API Change Methods in

`neighbors.NearestNeighbors`

:`kneighbors`

,`radius_neighbors`

,`kneighbors_graph`

,`radius_neighbors_graph`

now raise`NotFittedError`

, rather than`AttributeError`

, when called before`fit`

#12279 by Krishna Sangeeth.

`sklearn.neural_network`

¶

Fix Fixed a bug in

`neural_network.MLPClassifier`

and`neural_network.MLPRegressor`

where the option`shuffle=False`

was being ignored. #12582 by Sam Waterbury.Fix Fixed a bug in

`neural_network.MLPClassifier`

where validation sets for early stopping were not sampled with stratification. In the multilabel case however, splits are still not stratified. #13164 by Nicolas Hug.

`sklearn.pipeline`

¶

Feature

`pipeline.Pipeline`

can now use indexing notation (e.g.`my_pipeline[0:-1]`

) to extract a subsequence of steps as another Pipeline instance. A Pipeline can also be indexed directly to extract a particular step (e.g.`my_pipeline['svc']`

), rather than accessing`named_steps`

. #2568 by Joel Nothman.Feature Added optional parameter

`verbose`

in`pipeline.Pipeline`

,`compose.ColumnTransformer`

and`pipeline.FeatureUnion`

and corresponding`make_`

helpers for showing progress and timing of each step. #11364 by Baze Petrushev, Karan Desai, Joel Nothman, and Thomas Fan.Enhancement

`pipeline.Pipeline`

now supports using`'passthrough'`

as a transformer, with the same effect as`None`

. #11144 by Thomas Fan.Enhancement

`pipeline.Pipeline`

implements`__len__`

and therefore`len(pipeline)`

returns the number of steps in the pipeline. #13439 by Lakshya KD.

`sklearn.preprocessing`

¶

Feature

`preprocessing.OneHotEncoder`

now supports dropping one feature per category with a new drop parameter. #12908 by Drew Johnston.Efficiency

`preprocessing.OneHotEncoder`

and`preprocessing.OrdinalEncoder`

now handle pandas DataFrames more efficiently. #13253 by @maikia.Efficiency Make

`preprocessing.MultiLabelBinarizer`

cache class mappings instead of calculating it every time on the fly. #12116 by Ekaterina Krivich and Joel Nothman.Efficiency

`preprocessing.PolynomialFeatures`

now supports compressed sparse row (CSR) matrices as input for degrees 2 and 3. This is typically much faster than the dense case as it scales with matrix density and expansion degree (on the order of density^degree), and is much, much faster than the compressed sparse column (CSC) case. #12197 by Andrew Nystrom.Efficiency Speed improvement in

`preprocessing.PolynomialFeatures`

, in the dense case. Also added a new parameter`order`

which controls output order for further speed performances. #12251 by Tom Dupre la Tour.Fix Fixed the calculation overflow when using a float16 dtype with

`preprocessing.StandardScaler`

. #13007 by Raffaello BaluyotFix Fixed a bug in

`preprocessing.QuantileTransformer`

and`preprocessing.quantile_transform`

to force n_quantiles to be at most equal to n_samples. Values of n_quantiles larger than n_samples were either useless or resulting in a wrong approximation of the cumulative distribution function estimator. #13333 by Albert Thomas.API Change The default value of

`copy`

in`preprocessing.quantile_transform`

will change from False to True in 0.23 in order to make it more consistent with the default`copy`

values of other functions in`preprocessing`

and prevent unexpected side effects by modifying the value of`X`

inplace. #13459 by Hunter McGushion.

`sklearn.svm`

¶

Fix Fixed an issue in

`svm.SVC.decision_function`

when`decision_function_shape='ovr'`

. The decision_function value of a given sample was different depending on whether the decision_function was evaluated on the sample alone or on a batch containing this same sample due to the scaling used in decision_function. #10440 by Jonathan Ohayon.

`sklearn.tree`

¶

Feature Decision Trees can now be plotted with matplotlib using

`tree.plot_tree`

without relying on the`dot`

library, removing a hard-to-install dependency. #8508 by Andreas Müller.Feature Decision Trees can now be exported in a human readable textual format using

`tree.export_text`

. #6261 by`Giuseppe Vettigli <JustGlowing>`

.Feature

`get_n_leaves()`

and`get_depth()`

have been added to`tree.BaseDecisionTree`

and consequently all estimators based on it, including`tree.DecisionTreeClassifier`

,`tree.DecisionTreeRegressor`

,`tree.ExtraTreeClassifier`

, and`tree.ExtraTreeRegressor`

. #12300 by Adrin Jalali.Fix Trees and forests did not previously

`predict`

multi-output classification targets with string labels, despite accepting them in`fit`

. #11458 by Mitar Milutinovic.Fix Fixed an issue with

`tree.BaseDecisionTree`

and consequently all estimators based on it, including`tree.DecisionTreeClassifier`

,`tree.DecisionTreeRegressor`

,`tree.ExtraTreeClassifier`

, and`tree.ExtraTreeRegressor`

, where they used to exceed the given`max_depth`

by 1 while expanding the tree if`max_leaf_nodes`

and`max_depth`

were both specified by the user. Please note that this also affects all ensemble methods using decision trees. #12344 by Adrin Jalali.

`sklearn.utils`

¶

Feature

`utils.resample`

now accepts a`stratify`

parameter for sampling according to class distributions. #13549 by Nicolas Hug.API Change Deprecated

`warn_on_dtype`

parameter from`utils.check_array`

and`utils.check_X_y`

. Added explicit warning for dtype conversion in`check_pairwise_arrays`

if the`metric`

being passed is a pairwise boolean metric. #13382 by Prathmesh Savale.

### Multiple modules¶

Major Feature The

`__repr__()`

method of all estimators (used when calling`print(estimator)`

) has been entirely re-written, building on Python’s pretty printing standard library. All parameters are printed by default, but this can be altered with the`print_changed_only`

option in`sklearn.set_config`

. #11705 by Nicolas Hug.Major Feature Add estimators tags: these are annotations of estimators that allow programmatic inspection of their capabilities, such as sparse matrix support, supported output types and supported methods. Estimator tags also determine the tests that are run on an estimator when

`check_estimator`

is called. Read more in the User Guide. #8022 by Andreas Müller.Efficiency Memory copies are avoided when casting arrays to a different dtype in multiple estimators. #11973 by Roman Yurchak.

Fix Fixed a bug in the implementation of the

`our_rand_r`

helper function that was not behaving consistently across platforms. #13422 by Madhura Parikh and Clément Doumouro.

### Miscellaneous¶

Enhancement Joblib is no longer vendored in scikit-learn, and becomes a dependency. Minimal supported version is joblib 0.11, however using version >= 0.13 is strongly recommended. #13531 by Roman Yurchak.

## Changes to estimator checks¶

These changes mostly affect library developers.

Add

`check_fit_idempotent`

to`check_estimator`

, which checks that when`fit`

is called twice with the same data, the ouput of`predict`

,`predict_proba`

,`transform`

, and`decision_function`

does not change. #12328 by Nicolas HugMany checks can now be disabled or configured with Estimator Tags. #8022 by Andreas Müller.

## Code and Documentation Contributors¶

Thanks to everyone who has contributed to the maintenance and improvement of the project since version 0.20, including:

adanhawth, Aditya Vyas, Adrin Jalali, Agamemnon Krasoulis, Albert Thomas, Alberto Torres, Alexandre Gramfort, amourav, Andrea Navarrete, Andreas Mueller, Andrew Nystrom, assiaben, Aurélien Bellet, Bartosz Michałowski, Bartosz Telenczuk, bauks, BenjaStudio, bertrandhaut, Bharat Raghunathan, brentfagan, Bryan Woods, Cat Chenal, Cheuk Ting Ho, Chris Choe, Christos Aridas, Clément Doumouro, Cole Smith, Connossor, Corey Levinson, Dan Ellis, Dan Stine, Danylo Baibak, daten-kieker, Denis Kataev, Didi Bar-Zev, Dillon Gardner, Dmitry Mottl, Dmitry Vukolov, Dougal J. Sutherland, Dowon, drewmjohnston, Dror Atariah, Edward J Brown, Ekaterina Krivich, Elizabeth Sander, Emmanuel Arias, Eric Chang, Eric Larson, Erich Schubert, esvhd, Falak, Feda Curic, Federico Caselli, Frank Hoang, Fibinse Xavier`, Finn O’Shea, Gabriel Marzinotto, Gabriel Vacaliuc, Gabriele Calvo, Gael Varoquaux, GauravAhlawat, Giuseppe Vettigli, Greg Gandenberger, Guillaume Fournier, Guillaume Lemaitre, Gustavo De Mari Pereira, Hanmin Qin, haroldfox, hhu-luqi, Hunter McGushion, Ian Sanders, JackLangerman, Jacopo Notarstefano, jakirkham, James Bourbeau, Jan Koch, Jan S, janvanrijn, Jarrod Millman, jdethurens, jeremiedbb, JF, joaak, Joan Massich, Joel Nothman, Jonathan Ohayon, Joris Van den Bossche, josephsalmon, Jérémie Méhault, Katrin Leinweber, ken, kms15, Koen, Kossori Aruku, Krishna Sangeeth, Kuai Yu, Kulbear, Kushal Chauhan, Kyle Jackson, Lakshya KD, Leandro Hermida, Lee Yi Jie Joel, Lily Xiong, Lisa Sarah Thomas, Loic Esteve, louib, luk-f-a, maikia, mail-liam, Manimaran, Manuel López-Ibáñez, Marc Torrellas, Marco Gaido, Marco Gorelli, MarcoGorelli, marineLM, Mark Hannel, Martin Gubri, Masstran, mathurinm, Matthew Roeschke, Max Copeland, melsyt, mferrari3, Mickaël Schoentgen, Ming Li, Mitar, Mohammad Aftab, Mohammed AbdelAal, Mohammed Ibraheem, Muhammad Hassaan Rafique, mwestt, Naoya Iijima, Nicholas Smith, Nicolas Goix, Nicolas Hug, Nikolay Shebanov, Oleksandr Pavlyk, Oliver Rausch, Olivier Grisel, Orestis, Osman, Owen Flanagan, Paul Paczuski, Pavel Soriano, pavlos kallis, Pawel Sendyk, peay, Peter, Peter Cock, Peter Hausamann, Peter Marko, Pierre Glaser, pierretallotte, Pim de Haan, Piotr Szymański, Prabakaran Kumaresshan, Pradeep Reddy Raamana, Prathmesh Savale, Pulkit Maloo, Quentin Batista, Radostin Stoyanov, Raf Baluyot, Rajdeep Dua, Ramil Nugmanov, Raúl García Calvo, Rebekah Kim, Reshama Shaikh, Rohan Lekhwani, Rohan Singh, Rohan Varma, Rohit Kapoor, Roman Feldbauer, Roman Yurchak, Romuald M, Roopam Sharma, Ryan, Rüdiger Busche, Sam Waterbury, Samuel O. Ronsin, SandroCasagrande, Scott Cole, Scott Lowe, Sebastian Raschka, Shangwu Yao, Shivam Kotwalia, Shiyu Duan, smarie, Sriharsha Hatwar, Stephen Hoover, Stephen Tierney, Stéphane Couvreur, surgan12, SylvainLan, TakingItCasual, Tashay Green, thibsej, Thomas Fan, Thomas J Fan, Thomas Moreau, Tom Dupré la Tour, Tommy, Tulio Casagrande, Umar Farouk Umar, Utkarsh Upadhyay, Vinayak Mehta, Vishaal Kapoor, Vivek Kumar, Vlad Niculae, vqean3, Wenhao Zhang, William de Vazelhes, xhan, Xing Han Lu, xinyuliu12, Yaroslav Halchenko, Zach Griffith, Zach Miller, Zayd Hammoudeh, Zhuyi Xue, Zijie (ZJ) Poh, ^__^