<div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr">PyPI now has source and binary releases for Scikit-learn 0.21rc2.<div><br></div><div>* Documentation at <a href="https://scikit-learn.org/0.21">https://scikit-learn.org/0.21</a></div><div>* Release Notes at <a href="https://scikit-learn.org/0.21/whats_new">https://scikit-learn.org/0.21/whats_new</a><br></div><div>* Download source or wheels at <a href="https://pypi.org/project/scikit-learn/0.21rc2/">https://pypi.org/project/scikit-learn/0.21rc2/</a></div><div><br></div><div>Please try out the software and help us edit the release notes before a final release.</div><div><br></div><div>Highlights include:</div><div>* neighbors.NeighborhoodComponentsAnalysis for supervised metric learning, which learns a weighted euclidean distance for k-nearest neighbors. <a href="https://scikit-learn.org/0.21/modules/neighbors.html#nca">https://scikit-learn.org/0.21/modules/neighbors.html#nca</a></div><div>* ensemble.HistGradientBoostingClassifier and ensemble.HistGradientBoostingRegressor: experimental implementations of efficient binned gradient boosting machines. <a href="https://scikit-learn.org/0.21/modules/ensemble.html#gradient-tree-boosting">https://scikit-learn.org/0.21/modules/ensemble.html#gradient-tree-boosting</a></div><div>* impute.IterativeImputer: a non-trivial approach to missing value imputation. <a href="https://scikit-learn.org/0.21/modules/impute.html#multivariate-feature-imputation">https://scikit-learn.org/0.21/modules/impute.html#multivariate-feature-imputation</a><br></div><div>* cluster.OPTICS: a new density-based clustering algorithm. <a href="https://scikit-learn.org/0.21/modules/clustering.html#optics">https://scikit-learn.org/0.21/modules/clustering.html#optics</a></div><div>* better printing of estimators as strings, with an option to hide default parameters for compactness: <a href="https://scikit-learn.org/0.21/auto_examples/plot_changed_only_pprint_parameter.html">https://scikit-learn.org/0.21/auto_examples/plot_changed_only_pprint_parameter.html</a></div><div>* for estimator and library developers: a way to tag your estimator so that it can be treated appropriately with check_estimator. <a href="https://scikit-learn.org/0.21/developers/contributing.html#estimator-tags">https://scikit-learn.org/0.21/developers/contributing.html#estimator-tags</a></div><div><br></div><div>There are many other enhancements and fixes listed in the release notes (<a href="https://scikit-learn.org/0.21/whats_new">https://scikit-learn.org/0.21/whats_new</a>).</div><div><br></div><div>Please note that Scikit-learn has new dependencies:</div><div>* joblib >= 0.11, which used to be vendored within Scikit-learn</div><div>* OpenMP, unless the environment variable SKLEARN_NO_OPENMP=1 when the code is compiled (and cythonized)</div><div><br></div><div>Happy Learning!</div><div><br></div><div>From the Scikit-learn core dev team.</div></div></div></div></div></div></div></div></div></div>