Version 1.2#

For a short description of the main highlights of the release, please refer to Release Highlights for scikit-learn 1.2.

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 documented – 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.

Version 1.2.2#

March 2023

Changelog#

`sklearn.base`#

Fix When set_output(transform="pandas"), base.TransformerMixin maintains the index if the transform output is already a DataFrame. #25747 by Thomas Fan.

`sklearn.calibration`#

Fix A deprecation warning is raised when using the base_estimator__ prefix to set parameters of the estimator used in calibration.CalibratedClassifierCV. #25477 by Tim Head.

`sklearn.cluster`#

Fix Fixed a bug in cluster.BisectingKMeans, preventing fit to randomly fail due to a permutation of the labels when running multiple inits. #25563 by Jérémie du Boisberranger.

`sklearn.compose`#

Fix Fixes a bug in compose.ColumnTransformer which now supports empty selection of columns when set_output(transform="pandas"). #25570 by Thomas Fan.

`sklearn.ensemble`#

Fix A deprecation warning is raised when using the base_estimator__ prefix to set parameters of the estimator used in ensemble.AdaBoostClassifier, ensemble.AdaBoostRegressor, ensemble.BaggingClassifier, and ensemble.BaggingRegressor. #25477 by Tim Head.

`sklearn.feature_selection`#

Fix Fixed a regression where a negative tol would not be accepted any more by feature_selection.SequentialFeatureSelector. #25664 by Jérémie du Boisberranger.

`sklearn.inspection`#

Fix Raise a more informative error message in inspection.partial_dependence when dealing with mixed data type categories that cannot be sorted by numpy.unique. This problem usually happen when categories are str and missing values are present using np.nan. #25774 by Guillaume Lemaitre.

`sklearn.isotonic`#

Fix Fixes a bug in isotonic.IsotonicRegression where isotonic.IsotonicRegression.predict would return a pandas DataFrame when the global configuration sets transform_output="pandas". #25500 by Guillaume Lemaitre.

`sklearn.preprocessing`#

Fix preprocessing.OneHotEncoder.drop_idx_ now properly references the dropped category in the categories_ attribute when there are infrequent categories. #25589 by Thomas Fan.
Fix preprocessing.OrdinalEncoder now correctly supports encoded_missing_value or unknown_value set to a categories’ cardinality when there is missing values in the training data. #25704 by Thomas Fan.

`sklearn.tree`#

Fix Fixed a regression in tree.DecisionTreeClassifier, tree.DecisionTreeRegressor, tree.ExtraTreeClassifier and tree.ExtraTreeRegressor where an error was no longer raised in version 1.2 when min_sample_split=1. #25744 by Jérémie du Boisberranger.

`sklearn.utils`#

Fix Fixes a bug in utils.check_array which now correctly performs non-finite validation with the Array API specification. #25619 by Thomas Fan.
Fix utils.multiclass.type_of_target can identify pandas nullable data types as classification targets. #25638 by Thomas Fan.

Version 1.2.1#

January 2023

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.

Fix The fitted components in decomposition.MiniBatchDictionaryLearning might differ. The online updates of the sufficient statistics now properly take the sizes of the batches into account. #25354 by Jérémie du Boisberranger.
Fix The categories_ attribute of preprocessing.OneHotEncoder now always contains an array of object`s when using predefined categories that are strings. Predefined categories encoded as bytes will no longer work with `X encoded as strings. #25174 by Tim Head.

Changes impacting all modules#

Fix Support pandas.Int64 dtyped y for classifiers and regressors. #25089 by Tim Head.
Fix Remove spurious warnings for estimators internally using neighbors search methods. #25129 by Julien Jerphanion.
Fix Fix a bug where the current configuration was ignored in estimators using n_jobs > 1. This bug was triggered for tasks dispatched by the auxiliary thread of joblib as sklearn.get_config used to access an empty thread local configuration instead of the configuration visible from the thread where joblib.Parallel was first called. #25363 by Guillaume Lemaitre.

Changelog#

`sklearn.base`#

Fix Fix a regression in BaseEstimator.__getstate__ that would prevent certain estimators to be pickled when using Python 3.11. #25188 by Benjamin Bossan.
Fix Inheriting from base.TransformerMixin will only wrap the transform method if the class defines transform itself. #25295 by Thomas Fan.

`sklearn.datasets`#

Fix Fixes an inconsistency in datasets.fetch_openml between liac-arff and pandas parser when a leading space is introduced after the delimiter. The ARFF specs requires to ignore the leading space. #25312 by Guillaume Lemaitre.
Fix Fixes a bug in datasets.fetch_openml when using parser="pandas" where single quote and backslash escape characters were not properly handled. #25511 by Guillaume Lemaitre.

`sklearn.decomposition`#

Fix Fixed a bug in decomposition.MiniBatchDictionaryLearning where the online updates of the sufficient statistics where not correct when calling partial_fit on batches of different sizes. #25354 by Jérémie du Boisberranger.
Fix decomposition.DictionaryLearning better supports readonly NumPy arrays. In particular, it better supports large datasets which are memory-mapped when it is used with coordinate descent algorithms (i.e. when fit_algorithm='cd'). #25172 by Julien Jerphanion.

`sklearn.ensemble`#

Fix ensemble.RandomForestClassifier, ensemble.RandomForestRegressor ensemble.ExtraTreesClassifier and ensemble.ExtraTreesRegressor now support sparse readonly datasets. #25341 by Julien Jerphanion

`sklearn.feature_extraction`#

Fix feature_extraction.FeatureHasher raises an informative error when the input is a list of strings. #25094 by Thomas Fan.

`sklearn.linear_model`#

Fix Fix a regression in linear_model.SGDClassifier and linear_model.SGDRegressor that makes them unusable with the verbose parameter set to a value greater than 0. #25250 by Jérémie Du Boisberranger.

`sklearn.manifold`#

Fix manifold.TSNE now works correctly when output type is set to pandas #25370 by Tim Head.

`sklearn.model_selection`#

Fix model_selection.cross_validate with multimetric scoring in case of some failing scorers the non-failing scorers now returns proper scores instead of error_score values. #23101 by András Simon and Thomas Fan.

`sklearn.neural_network`#

Fix neural_network.MLPClassifier and neural_network.MLPRegressor no longer raise warnings when fitting data with feature names. #24873 by Tim Head.
Fix Improves error message in neural_network.MLPClassifier and neural_network.MLPRegressor, when early_stopping=True and partial_fit is called. #25694 by Thomas Fan.

`sklearn.preprocessing`#

Fix preprocessing.FunctionTransformer.inverse_transform correctly supports DataFrames that are all numerical when check_inverse=True. #25274 by Thomas Fan.
Fix preprocessing.SplineTransformer.get_feature_names_out correctly returns feature names when extrapolations="periodic". #25296 by Thomas Fan.

`sklearn.tree`#

Fix tree.DecisionTreeClassifier, tree.DecisionTreeRegressor tree.ExtraTreeClassifier and tree.ExtraTreeRegressor now support sparse readonly datasets. #25341 by Julien Jerphanion

`sklearn.utils`#

Fix Restore utils.check_array’s behaviour for pandas Series of type boolean. The type is maintained, instead of converting to float64. #25147 by Tim Head.
API Change utils.fixes.delayed is deprecated in 1.2.1 and will be removed in 1.5. Instead, import utils.parallel.delayed and use it in conjunction with the newly introduced utils.parallel.Parallel to ensure proper propagation of the scikit-learn configuration to the workers. #25363 by Guillaume Lemaitre.

Version 1.2.0#

December 2022

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.

Enhancement The default eigen_tol for cluster.SpectralClustering, manifold.SpectralEmbedding, cluster.spectral_clustering, and manifold.spectral_embedding is now None when using the 'amg' or 'lobpcg' solvers. This change improves numerical stability of the solver, but may result in a different model.
Enhancement linear_model.GammaRegressor, linear_model.PoissonRegressor and linear_model.TweedieRegressor can reach higher precision with the lbfgs solver, in particular when tol is set to a tiny value. Moreover, verbose is now properly propagated to L-BFGS-B. #23619 by Christian Lorentzen.
Enhancement The default value for eps metrics.log_loss has changed from 1e-15 to "auto". "auto" sets eps to np.finfo(y_pred.dtype).eps. #24354 by Safiuddin Khaja and gsiisg.
Fix Make sign of components_ deterministic in decomposition.SparsePCA. #23935 by Guillaume Lemaitre.
Fix The components_ signs in decomposition.FastICA might differ. It is now consistent and deterministic with all SVD solvers. #22527 by Meekail Zain and Thomas Fan.
Fix The condition for early stopping has now been changed in linear_model._sgd_fast._plain_sgd which is used by linear_model.SGDRegressor and linear_model.SGDClassifier. The old condition did not disambiguate between training and validation set and had an effect of overscaling the error tolerance. This has been fixed in #23798 by Harsh Agrawal.
Fix For model_selection.GridSearchCV and model_selection.RandomizedSearchCV ranks corresponding to nan scores will all be set to the maximum possible rank. #24543 by Guillaume Lemaitre.
API Change The default value of tol was changed from 1e-3 to 1e-4 for linear_model.ridge_regression, linear_model.Ridge and linear_model.RidgeClassifier. #24465 by Christian Lorentzen.

Changes impacting all modules#

Major Feature The set_output API has been adopted by all transformers. Meta-estimators that contain transformers such as pipeline.Pipeline or compose.ColumnTransformer also define a set_output. For details, see SLEP018. #23734 and #24699 by Thomas Fan.
Efficiency Low-level routines for reductions on pairwise distances for dense float32 datasets have been refactored. The following functions and estimators now benefit from improved performances in terms of hardware scalability and speed-ups:
For instance sklearn.neighbors.NearestNeighbors.kneighbors and sklearn.neighbors.NearestNeighbors.radius_neighbors can respectively be up to ×20 and ×5 faster than previously on a laptop.

Moreover, implementations of those two algorithms are now suitable for machine with many cores, making them usable for datasets consisting of millions of samples.

#23865 by Julien Jerphanion.
Enhancement Finiteness checks (detection of NaN and infinite values) in all estimators are now significantly more efficient for float32 data by leveraging NumPy’s SIMD optimized primitives. #23446 by Meekail Zain
Enhancement Finiteness checks (detection of NaN and infinite values) in all estimators are now faster by utilizing a more efficient stop-on-first second-pass algorithm. #23197 by Meekail Zain
Enhancement Support for combinations of dense and sparse datasets pairs for all distance metrics and for float32 and float64 datasets has been added or has seen its performance improved for the following estimators:
#23604 and #23585 by Julien Jerphanion, Olivier Grisel, and Thomas Fan, #24556 by Vincent Maladière.
Fix Systematically check the sha256 digest of dataset tarballs used in code examples in the documentation. #24617 by Olivier Grisel and Thomas Fan. Thanks to Sim4n6 for the report.

Changelog#

`sklearn.base`#

Enhancement Introduces base.ClassNamePrefixFeaturesOutMixin and base.ClassNamePrefixFeaturesOutMixin mixins that defines get_feature_names_out for common transformer uses cases. #24688 by Thomas Fan.

`sklearn.calibration`#

API Change Rename base_estimator to estimator in calibration.CalibratedClassifierCV to improve readability and consistency. The parameter base_estimator is deprecated and will be removed in 1.4. #22054 by Kevin Roice.

`sklearn.cluster`#

Efficiency cluster.KMeans with algorithm="lloyd" is now faster and uses less memory. #24264 by Vincent Maladiere.
Enhancement The predict and fit_predict methods of cluster.OPTICS now accept sparse data type for input data. #14736 by Hunt Zhan, #20802 by Brandon Pokorny, and #22965 by Meekail Zain.
Enhancement cluster.Birch now preserves dtype for numpy.float32 inputs. #22968 by Meekail Zain <micky774>.
Enhancement cluster.KMeans and cluster.MiniBatchKMeans now accept a new 'auto' option for n_init which changes the number of random initializations to one when using init='k-means++' for efficiency. This begins deprecation for the default values of n_init in the two classes and both will have their defaults changed to n_init='auto' in 1.4. #23038 by Meekail Zain.
Enhancement cluster.SpectralClustering and cluster.spectral_clustering now propagates the eigen_tol parameter to all choices of eigen_solver. Includes a new option eigen_tol="auto" and begins deprecation to change the default from eigen_tol=0 to eigen_tol="auto" in version 1.3. #23210 by Meekail Zain.
Fix cluster.KMeans now supports readonly attributes when predicting. #24258 by Thomas Fan
API Change The affinity attribute is now deprecated for cluster.AgglomerativeClustering and will be renamed to metric in v1.4. #23470 by Meekail Zain.

`sklearn.datasets`#

Enhancement Introduce the new parameter parser in datasets.fetch_openml. parser="pandas" allows to use the very CPU and memory efficient pandas.read_csv parser to load dense ARFF formatted dataset files. It is possible to pass parser="liac-arff" to use the old LIAC parser. When parser="auto", dense datasets are loaded with “pandas” and sparse datasets are loaded with “liac-arff”. Currently, parser="liac-arff" by default and will change to parser="auto" in version 1.4 #21938 by Guillaume Lemaitre.
Enhancement datasets.dump_svmlight_file is now accelerated with a Cython implementation, providing 2-4x speedups. #23127 by Meekail Zain
Enhancement Path-like objects, such as those created with pathlib are now allowed as paths in datasets.load_svmlight_file and datasets.load_svmlight_files. #19075 by Carlos Ramos Carreño.
Fix Make sure that datasets.fetch_lfw_people and datasets.fetch_lfw_pairs internally crops images based on the slice_ parameter. #24951 by Guillaume Lemaitre.

`sklearn.decomposition`#

Efficiency decomposition.FastICA.fit has been optimised w.r.t its memory footprint and runtime. #22268 by MohamedBsh.
Enhancement decomposition.SparsePCA and decomposition.MiniBatchSparsePCA now implements an inverse_transform function. #23905 by Guillaume Lemaitre.
Enhancement decomposition.FastICA now allows the user to select how whitening is performed through the new whiten_solver parameter, which supports svd and eigh. whiten_solver defaults to svd although eigh may be faster and more memory efficient in cases where num_features > num_samples. #11860 by Pierre Ablin, #22527 by Meekail Zain and Thomas Fan.
Enhancement decomposition.LatentDirichletAllocation now preserves dtype for numpy.float32 input. #24528 by Takeshi Oura and Jérémie du Boisberranger.
Fix Make sign of components_ deterministic in decomposition.SparsePCA. #23935 by Guillaume Lemaitre.
API Change The n_iter parameter of decomposition.MiniBatchSparsePCA is deprecated and replaced by the parameters max_iter, tol, and max_no_improvement to be consistent with decomposition.MiniBatchDictionaryLearning. n_iter will be removed in version 1.3. #23726 by Guillaume Lemaitre.
API Change The n_features_ attribute of decomposition.PCA is deprecated in favor of n_features_in_ and will be removed in 1.4. #24421 by Kshitij Mathur.

`sklearn.discriminant_analysis`#

Major Feature discriminant_analysis.LinearDiscriminantAnalysis now supports the Array API for solver="svd". Array API support is considered experimental and might evolve without being subjected to our usual rolling deprecation cycle policy. See Array API support (experimental) for more details. #22554 by Thomas Fan.
Fix Validate parameters only in fit and not in __init__ for discriminant_analysis.QuadraticDiscriminantAnalysis. #24218 by Stefanie Molin.

`sklearn.ensemble`#

Major Feature ensemble.HistGradientBoostingClassifier and ensemble.HistGradientBoostingRegressor now support interaction constraints via the argument interaction_cst of their constructors. #21020 by Christian Lorentzen. Using interaction constraints also makes fitting faster. #24856 by Christian Lorentzen.
Feature Adds class_weight to ensemble.HistGradientBoostingClassifier. #22014 by Thomas Fan.
Efficiency Improve runtime performance of ensemble.IsolationForest by avoiding data copies. #23252 by Zhehao Liu.
Enhancement ensemble.StackingClassifier now accepts any kind of base estimator. #24538 by Guillem G Subies.
Enhancement Make it possible to pass the categorical_features parameter of ensemble.HistGradientBoostingClassifier and ensemble.HistGradientBoostingRegressor as feature names. #24889 by Olivier Grisel.
Enhancement ensemble.StackingClassifier now supports multilabel-indicator target #24146 by Nicolas Peretti, Nestor Navarro, Nati Tomattis, and Vincent Maladiere.
Enhancement ensemble.HistGradientBoostingClassifier and ensemble.HistGradientBoostingClassifier now accept their monotonic_cst parameter to be passed as a dictionary in addition to the previously supported array-like format. Such dictionary have feature names as keys and one of -1, 0, 1 as value to specify monotonicity constraints for each feature. #24855 by Olivier Grisel.
Enhancement Interaction constraints for ensemble.HistGradientBoostingClassifier and ensemble.HistGradientBoostingRegressor can now be specified as strings for two common cases: “no_interactions” and “pairwise” interactions. #24849 by Tim Head.
Fix Fixed the issue where ensemble.AdaBoostClassifier outputs NaN in feature importance when fitted with very small sample weight. #20415 by Zhehao Liu.
Fix ensemble.HistGradientBoostingClassifier and ensemble.HistGradientBoostingRegressor no longer error when predicting on categories encoded as negative values and instead consider them a member of the “missing category”. #24283 by Thomas Fan.
Fix ensemble.HistGradientBoostingClassifier and ensemble.HistGradientBoostingRegressor, with verbose>=1, print detailed timing information on computing histograms and finding best splits. The time spent in the root node was previously missing and is now included in the printed information. #24894 by Christian Lorentzen.
API Change Rename the constructor parameter base_estimator to estimator in the following classes: ensemble.BaggingClassifier, ensemble.BaggingRegressor, ensemble.AdaBoostClassifier, ensemble.AdaBoostRegressor. base_estimator is deprecated in 1.2 and will be removed in 1.4. #23819 by Adrian Trujillo and Edoardo Abati.
API Change Rename the fitted attribute base_estimator_ to estimator_ in the following classes: ensemble.BaggingClassifier, ensemble.BaggingRegressor, ensemble.AdaBoostClassifier, ensemble.AdaBoostRegressor, ensemble.RandomForestClassifier, ensemble.RandomForestRegressor, ensemble.ExtraTreesClassifier, ensemble.ExtraTreesRegressor, ensemble.RandomTreesEmbedding, ensemble.IsolationForest. base_estimator_ is deprecated in 1.2 and will be removed in 1.4. #23819 by Adrian Trujillo and Edoardo Abati.

`sklearn.feature_selection`#

Fix Fix a bug in feature_selection.mutual_info_regression and feature_selection.mutual_info_classif, where the continuous features in X should be scaled to a unit variance independently if the target y is continuous or discrete. #24747 by Guillaume Lemaitre

`sklearn.gaussian_process`#

Fix Fix gaussian_process.kernels.Matern gradient computation with nu=0.5 for PyPy (and possibly other non CPython interpreters). #24245 by Loïc Estève.
Fix The fit method of gaussian_process.GaussianProcessRegressor will not modify the input X in case a custom kernel is used, with a diag method that returns part of the input X. #24405 by Omar Salman.

`sklearn.impute`#

Enhancement Added keep_empty_features parameter to impute.SimpleImputer, impute.KNNImputer and impute.IterativeImputer, preventing removal of features containing only missing values when transforming. #16695 by Vitor Santa Rosa.

`sklearn.inspection`#

Major Feature Extended inspection.partial_dependence and inspection.PartialDependenceDisplay to handle categorical features. #18298 by Madhura Jayaratne and Guillaume Lemaitre.
Fix inspection.DecisionBoundaryDisplay now raises error if input data is not 2-dimensional. #25077 by Arturo Amor.

`sklearn.kernel_approximation`#

Enhancement kernel_approximation.RBFSampler now preserves dtype for numpy.float32 inputs. #24317 by Tim Head <betatim>.
Enhancement kernel_approximation.SkewedChi2Sampler now preserves dtype for numpy.float32 inputs. #24350 by Rahil Parikh.
Enhancement kernel_approximation.RBFSampler now accepts 'scale' option for parameter gamma. #24755 by Gleb Levitski.

`sklearn.linear_model`#

Enhancement linear_model.LogisticRegression, linear_model.LogisticRegressionCV, linear_model.GammaRegressor, linear_model.PoissonRegressor and linear_model.TweedieRegressor got a new solver solver="newton-cholesky". This is a 2nd order (Newton) optimisation routine that uses a Cholesky decomposition of the hessian matrix. When n_samples >> n_features, the "newton-cholesky" solver has been observed to converge both faster and to a higher precision solution than the "lbfgs" solver on problems with one-hot encoded categorical variables with some rare categorical levels. #24637 and #24767 by Christian Lorentzen.
Enhancement linear_model.GammaRegressor, linear_model.PoissonRegressor and linear_model.TweedieRegressor can reach higher precision with the lbfgs solver, in particular when tol is set to a tiny value. Moreover, verbose is now properly propagated to L-BFGS-B. #23619 by Christian Lorentzen.
Fix linear_model.SGDClassifier and linear_model.SGDRegressor will raise an error when all the validation samples have zero sample weight. #23275 by Zhehao Liu <MaxwellLZH>.
Fix linear_model.SGDOneClassSVM no longer performs parameter validation in the constructor. All validation is now handled in fit() and partial_fit(). #24433 by Yogendrasingh, Arisa Y. and Tim Head.
Fix Fix average loss calculation when early stopping is enabled in linear_model.SGDRegressor and linear_model.SGDClassifier. Also updated the condition for early stopping accordingly. #23798 by Harsh Agrawal.
API Change The default value for the solver parameter in linear_model.QuantileRegressor will change from "interior-point" to "highs" in version 1.4. #23637 by Guillaume Lemaitre.
API Change String option "none" is deprecated for penalty argument in linear_model.LogisticRegression, and will be removed in version 1.4. Use None instead. #23877 by Zhehao Liu.
API Change The default value of tol was changed from 1e-3 to 1e-4 for linear_model.ridge_regression, linear_model.Ridge and linear_model.RidgeClassifier. #24465 by Christian Lorentzen.

`sklearn.manifold`#

Feature Adds option to use the normalized stress in manifold.MDS. This is enabled by setting the new normalize parameter to True. #10168 by Łukasz Borchmann, #12285 by Matthias Miltenberger, #13042 by Matthieu Parizy, #18094 by Roth E Conrad and #22562 by Meekail Zain.
Enhancement Adds eigen_tol parameter to manifold.SpectralEmbedding. Both manifold.spectral_embedding and manifold.SpectralEmbedding now propagate eigen_tol to all choices of eigen_solver. Includes a new option eigen_tol="auto" and begins deprecation to change the default from eigen_tol=0 to eigen_tol="auto" in version 1.3. #23210 by Meekail Zain.
Enhancement manifold.Isomap now preserves dtype for np.float32 inputs. #24714 by Rahil Parikh.
API Change Added an "auto" option to the normalized_stress argument in manifold.MDS and manifold.smacof. Note that normalized_stress is only valid for non-metric MDS, therefore the "auto" option enables normalized_stress when metric=False and disables it when metric=True. "auto" will become the default value for normalized_stress in version 1.4. #23834 by Meekail Zain

`sklearn.metrics`#

Feature metrics.ConfusionMatrixDisplay.from_estimator, metrics.ConfusionMatrixDisplay.from_predictions, and metrics.ConfusionMatrixDisplay.plot accepts a text_kw parameter which is passed to matplotlib’s text function. #24051 by Thomas Fan.
Feature metrics.class_likelihood_ratios is added to compute the positive and negative likelihood ratios derived from the confusion matrix of a binary classification problem. #22518 by Arturo Amor.
Feature Add metrics.PredictionErrorDisplay to plot residuals vs predicted and actual vs predicted to qualitatively assess the behavior of a regressor. The display can be created with the class methods metrics.PredictionErrorDisplay.from_estimator and metrics.PredictionErrorDisplay.from_predictions. #18020 by Guillaume Lemaitre.
Feature metrics.roc_auc_score now supports micro-averaging (average="micro") for the One-vs-Rest multiclass case (multi_class="ovr"). #24338 by Arturo Amor.
Enhancement Adds an "auto" option to eps in metrics.log_loss. This option will automatically set the eps value depending on the data type of y_pred. In addition, the default value of eps is changed from 1e-15 to the new "auto" option. #24354 by Safiuddin Khaja and gsiisg.
Fix Allows csr_matrix as input for parameter: y_true of the metrics.label_ranking_average_precision_score metric. #23442 by Sean Atukorala
Fix metrics.ndcg_score will now trigger a warning when the y_true value contains a negative value. Users may still use negative values, but the result may not be between 0 and 1. Starting in v1.4, passing in negative values for y_true will raise an error. #22710 by Conroy Trinh and #23461 by Meekail Zain.
Fix metrics.log_loss with eps=0 now returns a correct value of 0 or np.inf instead of nan for predictions at the boundaries (0 or 1). It also accepts integer input. #24365 by Christian Lorentzen.
API Change The parameter sum_over_features of metrics.pairwise.manhattan_distances is deprecated and will be removed in 1.4. #24630 by Rushil Desai.

`sklearn.model_selection`#

Feature Added the class model_selection.LearningCurveDisplay that allows to make easy plotting of learning curves obtained by the function model_selection.learning_curve. #24084 by Guillaume Lemaitre.
Fix For all SearchCV classes and scipy >= 1.10, rank corresponding to a nan score is correctly set to the maximum possible rank, rather than np.iinfo(np.int32).min. #24141 by Loïc Estève.
Fix In both model_selection.HalvingGridSearchCV and model_selection.HalvingRandomSearchCV parameter combinations with a NaN score now share the lowest rank. #24539 by Tim Head.
Fix For model_selection.GridSearchCV and model_selection.RandomizedSearchCV ranks corresponding to nan scores will all be set to the maximum possible rank. #24543 by Guillaume Lemaitre.

`sklearn.multioutput`#

Feature Added boolean verbose flag to classes: multioutput.ClassifierChain and multioutput.RegressorChain. #23977 by Eric Fiegel, Chiara Marmo, Lucy Liu, and Guillaume Lemaitre.

`sklearn.naive_bayes`#

Feature Add methods predict_joint_log_proba to all naive Bayes classifiers. #23683 by Andrey Melnik.
Enhancement A new parameter force_alpha was added to naive_bayes.BernoulliNB, naive_bayes.ComplementNB, naive_bayes.CategoricalNB, and naive_bayes.MultinomialNB, allowing user to set parameter alpha to a very small number, greater or equal 0, which was earlier automatically changed to 1e-10 instead. #16747 by @arka204, #18805 by @hongshaoyang, #22269 by Meekail Zain.

`sklearn.neighbors`#

Feature Adds new function neighbors.sort_graph_by_row_values to sort a CSR sparse graph such that each row is stored with increasing values. This is useful to improve efficiency when using precomputed sparse distance matrices in a variety of estimators and avoid an EfficiencyWarning. #23139 by Tom Dupre la Tour.
Efficiency neighbors.NearestCentroid is faster and requires less memory as it better leverages CPUs’ caches to compute predictions. #24645 by Olivier Grisel.
Enhancement neighbors.KernelDensity bandwidth parameter now accepts definition using Scott’s and Silverman’s estimation methods. #10468 by Ruben and #22993 by Jovan Stojanovic.
Enhancement neighbors.NeighborsBase now accepts Minkowski semi-metric (i.e. when \(0 < p < 1\) for metric="minkowski") for algorithm="auto" or algorithm="brute". #24750 by Rudresh Veerkhare
Fix neighbors.NearestCentroid now raises an informative error message at fit-time instead of failing with a low-level error message at predict-time. #23874 by Juan Gomez.
Fix Set n_jobs=None by default (instead of 1) for neighbors.KNeighborsTransformer and neighbors.RadiusNeighborsTransformer. #24075 by Valentin Laurent.
Enhancement neighbors.LocalOutlierFactor now preserves dtype for numpy.float32 inputs. #22665 by Julien Jerphanion.

`sklearn.neural_network`#

Fix neural_network.MLPClassifier and neural_network.MLPRegressor always expose the parameters best_loss_, validation_scores_, and best_validation_score_. best_loss_ is set to None when early_stopping=True, while validation_scores_ and best_validation_score_ are set to None when early_stopping=False. #24683 by Guillaume Lemaitre.

`sklearn.pipeline`#

Enhancement pipeline.FeatureUnion.get_feature_names_out can now be used when one of the transformers in the pipeline.FeatureUnion is "passthrough". #24058 by Diederik Perdok
Enhancement The pipeline.FeatureUnion class now has a named_transformers attribute for accessing transformers by name. #20331 by Christopher Flynn.

`sklearn.preprocessing`#

Enhancement preprocessing.FunctionTransformer will always try to set n_features_in_ and feature_names_in_ regardless of the validate parameter. #23993 by Thomas Fan.
Fix preprocessing.LabelEncoder correctly encodes NaNs in transform. #22629 by Thomas Fan.
API Change The sparse parameter of preprocessing.OneHotEncoder is now deprecated and will be removed in version 1.4. Use sparse_output instead. #24412 by Rushil Desai.

`sklearn.svm`#

API Change The class_weight_ attribute is now deprecated for svm.NuSVR, svm.SVR, svm.OneClassSVM. #22898 by Meekail Zain.

`sklearn.tree`#

Enhancement tree.plot_tree, tree.export_graphviz now uses a lower case x[i] to represent feature i. #23480 by Thomas Fan.

`sklearn.utils`#

Feature A new module exposes development tools to discover estimators (i.e. utils.discovery.all_estimators), displays (i.e. utils.discovery.all_displays) and functions (i.e. utils.discovery.all_functions) in scikit-learn. #21469 by Guillaume Lemaitre.
Enhancement utils.extmath.randomized_svd now accepts an argument, lapack_svd_driver, to specify the lapack driver used in the internal deterministic SVD used by the randomized SVD algorithm. #20617 by Srinath Kailasa
Enhancement utils.validation.column_or_1d now accepts a dtype parameter to specific y’s dtype. #22629 by Thomas Fan.
Enhancement utils.extmath.cartesian now accepts arrays with different dtype and will cast the output to the most permissive dtype. #25067 by Guillaume Lemaitre.
Fix utils.multiclass.type_of_target now properly handles sparse matrices. #14862 by Léonard Binet.
Fix HTML representation no longer errors when an estimator class is a value in get_params. #24512 by Thomas Fan.
Fix utils.estimator_checks.check_estimator now takes into account the requires_positive_X tag correctly. #24667 by Thomas Fan.
Fix utils.check_array now supports Pandas Series with pd.NA by raising a better error message or returning a compatible ndarray. #25080 by Thomas Fan.
API Change The extra keyword parameters of utils.extmath.density are deprecated and will be removed in 1.4. #24523 by Mia Bajic.

Code and documentation contributors

Thanks to everyone who has contributed to the maintenance and improvement of the project since version 1.1, including:

2357juan, 3lLobo, Adam J. Stewart, Adam Kania, Adam Li, Aditya Anulekh, Admir Demiraj, adoublet, Adrin Jalali, Ahmedbgh, Aiko, Akshita Prasanth, Ala-Na, Alessandro Miola, Alex, Alexandr, Alexandre Perez-Lebel, Alex Buzenet, Ali H. El-Kassas, aman kumar, Amit Bera, András Simon, Andreas Grivas, Andreas Mueller, Andrew Wang, angela-maennel, Aniket Shirsat, Anthony22-dev, Antony Lee, anupam, Apostolos Tsetoglou, Aravindh R, Artur Hermano, Arturo Amor, as-90, ashah002, Ashwin Mathur, avm19, Azaria Gebremichael, b0rxington, Badr MOUFAD, Bardiya Ak, Bartłomiej Gońda, BdeGraaff, Benjamin Bossan, Benjamin Carter, berkecanrizai, Bernd Fritzke, Bhoomika, Biswaroop Mitra, Brandon TH Chen, Brett Cannon, Bsh, cache-missing, carlo, Carlos Ramos Carreño, ceh, chalulu, Changyao Chen, Charles Zablit, Chiara Marmo, Christian Lorentzen, Christian Ritter, Christian Veenhuis, christianwaldmann, Christine P. Chai, Claudio Salvatore Arcidiacono, Clément Verrier, crispinlogan, Da-Lan, DanGonite57, Daniela Fernandes, DanielGaerber, darioka, Darren Nguyen, davidblnc, david-cortes, David Gilbertson, David Poznik, Dayne, Dea María Léon, Denis, Dev Khant, Dhanshree Arora, Diadochokinetic, diederikwp, Dimitri Papadopoulos Orfanos, Dimitris Litsidis, drewhogg, Duarte OC, Dwight Lindquist, Eden Brekke, Edern, Edoardo Abati, Eleanore Denies, EliaSchiavon, Emir, ErmolaevPA, Fabrizio Damicelli, fcharras, Felipe Siola, Flynn, francesco-tuveri, Franck Charras, ftorres16, Gael Varoquaux, Geevarghese George, genvalen, GeorgiaMayDay, Gianr Lazz, Gleb Levitski, Glòria Macià Muñoz, Guillaume Lemaitre, Guillem García Subies, Guitared, gunesbayir, Haesun Park, Hansin Ahuja, Hao Chun Chang, Harsh Agrawal, harshit5674, hasan-yaman, henrymooresc, Henry Sorsky, Hristo Vrigazov, htsedebenham, humahn, i-aki-y, Ian Thompson, Ido M, Iglesys, Iliya Zhechev, Irene, ivanllt, Ivan Sedykh, Jack McIvor, jakirkham, JanFidor, Jason G, Jérémie du Boisberranger, Jiten Sidhpura, jkarolczak, João David, JohnathanPi, John Koumentis, John P, John Pangas, johnthagen, Jordan Fleming, Joshua Choo Yun Keat, Jovan Stojanovic, Juan Carlos Alfaro Jiménez, juanfe88, Juan Felipe Arias, JuliaSchoepp, Julien Jerphanion, jygerardy, ka00ri, Kanishk Sachdev, Kanissh, Kaushik Amar Das, Kendall, Kenneth Prabakaran, Kento Nozawa, kernc, Kevin Roice, Kian Eliasi, Kilian Kluge, Kilian Lieret, Kirandevraj, Kraig, krishna kumar, krishna vamsi, Kshitij Kapadni, Kshitij Mathur, Lauren Burke, Léonard Binet, lingyi1110, Lisa Casino, Logan Thomas, Loic Esteve, Luciano Mantovani, Lucy Liu, Maascha, Madhura Jayaratne, madinak, Maksym, Malte S. Kurz, Mansi Agrawal, Marco Edward Gorelli, Marco Wurps, Maren Westermann, Maria Telenczuk, Mario Kostelac, martin-kokos, Marvin Krawutschke, Masanori Kanazu, mathurinm, Matt Haberland, mauroantonioserrano, Max Halford, Maxi Marufo, maximeSaur, Maxim Smolskiy, Maxwell, m. bou, Meekail Zain, Mehgarg, mehmetcanakbay, Mia Bajić, Michael Flaks, Michael Hornstein, Michel de Ruiter, Michelle Paradis, Mikhail Iljin, Misa Ogura, Moritz Wilksch, mrastgoo, Naipawat Poolsawat, Naoise Holohan, Nass, Nathan Jacobi, Nawazish Alam, Nguyễn Văn Diễn, Nicola Fanelli, Nihal Thukarama Rao, Nikita Jare, nima10khodaveisi, Nima Sarajpoor, nitinramvelraj, NNLNR, npache, Nwanna-Joseph, Nymark Kho, o-holman, Olivier Grisel, Olle Lukowski, Omar Hassoun, Omar Salman, osman tamer, ouss1508, Oyindamola Olatunji, PAB, Pandata, partev, Paulo Sergio Soares, Petar Mlinarić, Peter Jansson, Peter Steinbach, Philipp Jung, Piet Brömmel, Pooja M, Pooja Subramaniam, priyam kakati, puhuk, Rachel Freeland, Rachit Keerti Das, Rafal Wojdyla, Raghuveer Bhat, Rahil Parikh, Ralf Gommers, ram vikram singh, Ravi Makhija, Rehan Guha, Reshama Shaikh, Richard Klima, Rob Crockett, Robert Hommes, Robert Juergens, Robin Lenz, Rocco Meli, Roman4oo, Ross Barnowski, Rowan Mankoo, Rudresh Veerkhare, Rushil Desai, Sabri Monaf Sabri, Safikh, Safiuddin Khaja, Salahuddin, Sam Adam Day, Sandra Yojana Meneses, Sandro Ephrem, Sangam, SangamSwadik, SANJAI_3, SarahRemus, Sashka Warner, SavkoMax, Scott Gigante, Scott Gustafson, Sean Atukorala, sec65, SELEE, seljaks, Shady el Gewily, Shane, shellyfung, Shinsuke Mori, Shiva chauhan, Shoaib Khan, Shogo Hida, Shrankhla Srivastava, Shuangchi He, Simon, sonnivs, Sortofamudkip, Srinath Kailasa, Stanislav (Stanley) Modrak, Stefanie Molin, stellalin7, Stéphane Collot, Steven Van Vaerenbergh, Steve Schmerler, Sven Stehle, Tabea Kossen, TheDevPanda, the-syd-sre, Thijs van Weezel, Thomas Bonald, Thomas Germer, Thomas J. Fan, Ti-Ion, Tim Head, Timofei Kornev, toastedyeast, Tobias Pitters, Tom Dupré la Tour, tomiock, Tom Mathews, Tom McTiernan, tspeng, Tyler Egashira, Valentin Laurent, Varun Jain, Vera Komeyer, Vicente Reyes-Puerta, Vinayak Mehta, Vincent M, Vishal, Vyom Pathak, wattai, wchathura, WEN Hao, William M, x110, Xiao Yuan, Xunius, yanhong-zhao-ef, Yusuf Raji, Z Adil Khwaja, zeeshan lone