Merge branch 'scikit-learn:main' into submodulev3

Author: adam2392

build_tools/azure/install.sh

@@ -49,7 +49,10 @@ pre_python_environment_install() {
 
 python_environment_install_and_activate() {
     if [[ "$DISTRIB" == "conda"* ]]; then
-        conda update -n base conda -y
+        # Install/update conda with the libmamba solver because the legacy
+        # solver can be slow at installing a specific version of conda-lock.
+        conda install -n base conda conda-libmamba-solver -y
+        conda config --set solver libmamba
         conda install -c conda-forge "$(get_dep conda-lock min)" -y
         conda-lock install --name $VIRTUALENV $LOCK_FILE
         source activate $VIRTUALENV

build_tools/azure/install_win.sh

@@ -7,6 +7,10 @@ set -x
 source build_tools/shared.sh
 
 if [[ "$DISTRIB" == "conda" ]]; then
+    # Install/update conda with the libmamba solver because the legacy solver
+    # can be slow at installing a specific version of conda-lock.
+    conda install -n base conda conda-libmamba-solver -y
+    conda config --set solver libmamba
     conda install -c conda-forge "$(get_dep conda-lock min)" -y
     conda-lock install --name $VIRTUALENV $LOCK_FILE
     source activate $VIRTUALENV

doc/conftest.py

@@ -145,13 +145,6 @@ def pytest_runtest_setup(item):
         setup_preprocessing()
     elif fname.endswith("statistical_inference/unsupervised_learning.rst"):
         setup_unsupervised_learning()
-    elif fname.endswith("metadata_routing.rst"):
-        # TODO: remove this once implemented
-        # Skip metarouting because is it is not fully implemented yet
-        raise SkipTest(
-            "Skipping doctest for metadata_routing.rst because it "
-            "is not fully implemented yet"
-        )
 
     rst_files_requiring_matplotlib = [
         "modules/partial_dependence.rst",

doc/developers/advanced_installation.rst

@@ -69,6 +69,12 @@ feature, code or documentation improvement).
    .. prompt:: bash $
 
      conda create -n sklearn-env -c conda-forge python=3.9 numpy scipy cython
+
+   It is not always necessary but it is safer to open a new prompt before
+   activating the newly created conda environment.
+
+   .. prompt:: bash $
+
      conda activate sklearn-env
 
 #. **Alternative to conda:** If you run Linux or similar, you can instead use
@@ -287,6 +293,12 @@ scikit-learn from source:
 
     conda create -n sklearn-dev -c conda-forge python numpy scipy cython \
         joblib threadpoolctl pytest compilers llvm-openmp
+
+It is not always necessary but it is safer to open a new prompt before
+activating the newly created conda environment.
+
+.. prompt:: bash $
+
     conda activate sklearn-dev
     make clean
     pip install -v --no-use-pep517 --no-build-isolation -e .
@@ -307,12 +319,6 @@ forge using the following command:
 
 which should include ``compilers`` and ``llvm-openmp``.
 
-.. note::
-
-   If you installed these packages after creating and activating a new conda
-   environment, you will need to first deactivate and then reactivate the
-   environment for these changes to take effect.
-
 The compilers meta-package will automatically set custom environment
 variables:
 
@@ -428,6 +434,12 @@ in the user folder using conda:
 
     conda create -n sklearn-dev -c conda-forge python numpy scipy cython \
         joblib threadpoolctl pytest compilers
+
+It is not always necessary but it is safer to open a new prompt before
+activating the newly created conda environment.
+
+.. prompt:: bash $
+
     conda activate sklearn-dev
     pip install -v --no-use-pep517 --no-build-isolation -e .
 

doc/metadata_routing.rst

@@ -76,15 +76,15 @@ metadata called ``sample_weight``::
   ...     lr,
   ...     X,
   ...     y,
-  ...     props={"sample_weight": my_weights, "groups": my_groups},
+  ...     params={"sample_weight": my_weights, "groups": my_groups},
   ...     cv=GroupKFold(),
   ...     scoring=weighted_acc,
   ... )
 
 Note that in this example, ``my_weights`` is passed to both the scorer and
 :class:`~linear_model.LogisticRegressionCV`.
 
-Error handling: if ``props={"sample_weigh": my_weights, ...}`` were passed
+Error handling: if ``params={"sample_weigh": my_weights, ...}`` were passed
 (note the typo), :func:`~model_selection.cross_validate` would raise an error,
 since ``sample_weigh`` was not requested by any of its underlying objects.
 
@@ -110,7 +110,7 @@ that :func:`~model_selection.cross_validate` does not pass the weights along::
   ...     X,
   ...     y,
   ...     cv=GroupKFold(),
-  ...     props={"sample_weight": my_weights, "groups": my_groups},
+  ...     params={"sample_weight": my_weights, "groups": my_groups},
   ...     scoring=weighted_acc,
   ... )
 
@@ -142,7 +142,7 @@ instance is set and ``sample_weight`` is not routed to it::
   ...     X,
   ...     y,
   ...     cv=GroupKFold(),
-  ...     props={"sample_weight": my_weights, "groups": my_groups},
+  ...     params={"sample_weight": my_weights, "groups": my_groups},
   ...     scoring=weighted_acc,
   ... )
 
@@ -166,7 +166,7 @@ consumers. In this example, we pass ``scoring_weight`` to the scorer, and
   ...     X,
   ...     y,
   ...     cv=GroupKFold(),
-  ...     props={
+  ...     params={
   ...         "scoring_weight": my_weights,
   ...         "fitting_weight": my_other_weights,
   ...         "groups": my_groups,

doc/modules/classes.rst

@@ -995,6 +995,8 @@ details.
    metrics.median_absolute_error
    metrics.mean_absolute_percentage_error
    metrics.r2_score
+   metrics.root_mean_squared_log_error
+   metrics.root_mean_squared_error
    metrics.mean_poisson_deviance
    metrics.mean_gamma_deviance
    metrics.mean_tweedie_deviance

doc/modules/model_evaluation.rst

@@ -94,8 +94,9 @@ Scoring                                Function
 'max_error'                            :func:`metrics.max_error`
 'neg_mean_absolute_error'              :func:`metrics.mean_absolute_error`
 'neg_mean_squared_error'               :func:`metrics.mean_squared_error`
-'neg_root_mean_squared_error'          :func:`metrics.mean_squared_error`
+'neg_root_mean_squared_error'          :func:`metrics.root_mean_squared_error`
 'neg_mean_squared_log_error'           :func:`metrics.mean_squared_log_error`
+'neg_root_mean_squared_log_error'      :func:`metrics.root_mean_squared_log_error`
 'neg_median_absolute_error'            :func:`metrics.median_absolute_error`
 'r2'                                   :func:`metrics.r2_score`
 'neg_mean_poisson_deviance'            :func:`metrics.mean_poisson_deviance`
@@ -2310,6 +2311,10 @@ function::
     for an example of mean squared error usage to
     evaluate gradient boosting regression.
 
+Taking the square root of the MSE, called the root mean squared error (RMSE), is another
+common metric that provides a measure in the same units as the target variable. RSME is
+available through the :func:`root_mean_squared_error` function.
+
 .. _mean_squared_log_error:
 
 Mean squared logarithmic error
@@ -2347,6 +2352,9 @@ function::
   >>> mean_squared_log_error(y_true, y_pred)
   0.044...
 
+The root mean squared logarithmic error (RMSLE) is available through the
+:func:`root_mean_squared_log_error` function.
+
 .. _mean_absolute_percentage_error:
 
 Mean absolute percentage error

doc/related_projects.rst

@@ -228,6 +228,14 @@ Note scikit-learn own modern gradient boosting estimators
 - `Flower <https://flower.dev/>`_ A friendly federated learning framework with a
   unified approach that can federate any workload, any ML framework, and any programming language.
 
+**Privacy Preserving Machine Learning**
+
+- `Concrete ML <https://github.com/zama-ai/concrete-ml/>`_ A privacy preserving
+  ML framework built on top of `Concrete
+  <https://github.com/zama-ai/concrete>`_, with bindings to traditional ML
+  frameworks, thanks to fully homomorphic encryption. APIs of so-called
+  Concrete ML built-in models are very close to scikit-learn APIs.
+
 **Broad scope**
 
 - `mlxtend <https://github.com/rasbt/mlxtend>`_ Includes a number of additional

doc/whats_new/v1.3.rst

@@ -9,6 +9,20 @@ Version 1.3.1
 
 **In development**
 
+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| Ridge models with `solver='sparse_cg'` may have slightly different
+  results with scipy>=1.12, because of an underlying change in the scipy solver
+  (see `scipy#18488 <https://github.com/scipy/scipy/pull/18488>`_ for more
+  details)
+  :pr:`26814` by :user:`Loïc Estève <lesteve>`
+
 Changes impacting all modules
 -----------------------------
 
@@ -18,6 +32,13 @@ Changes impacting all modules
 Changelog
 ---------
 
+:mod:`sklearn.calibration`
+..........................
+
+- |Fix| :class:`calibration.CalibratedClassifierCV` can now handle models that
+  produce large prediction scores. Before it was numerically unstable.
+  :pr:`26913` by :user:`Omar Salman <OmarManzoor>`.
+
 :mod:`sklearn.cluster`
 ......................
 
@@ -26,7 +47,14 @@ Changelog
   :pr:`27167` by `Olivier Grisel`_.
 
 - |Fix| :class:`cluster.BisectingKMeans` now works with data that has a single feature.
-  :pr:`27243` by `Jérémie du Boisberranger <jeremiedbb>`.
+  :pr:`27243` by :user:`Jérémie du Boisberranger <jeremiedbb>`.
+
+:mod:`sklearn.cross_decomposition`
+..................................
+
+- |Fix| :class:`cross_decomposition.PLSRegression` now automatically ravels the output
+  of `predict` if fitted with one dimensional `y`.
+  :pr:`26602` by :user:`Yao Xiao <Charlie-XIAO>`.
 
 :mod:`sklearn.ensemble`
 .......................
@@ -36,13 +64,34 @@ Changelog
   the sum of the scores should sum to zero for a sample).
   :pr:`26521` by :user:`Guillaume Lemaitre <glemaitre>`.
 
+:mod:`sklearn.feature_selection`
+................................
+
+- |Fix| :func:`feature_selection.mutual_info_regression` now correctly computes the
+  result when `X` is of integer dtype. :pr:`26748` by :user:`Yao Xiao <Charlie-XIAO>`.
+
 :mod:`sklearn.impute`
 .....................
 
 - |Fix| :class:`impute.KNNImputer` now correctly adds a missing indicator column in
   ``transform`` when ``add_indicator`` is set to ``True`` and missing values are observed
   during ``fit``. :pr:`26600` by :user:`Shreesha Kumar Bhat <Shreesha3112>`.
 
+:mod:`sklearn.metrics`
+......................
+
+- |Fix| Scorers used with :func:`metrics.get_scorer` handle properly
+  multilabel-indicator matrix.
+  :pr:`27002` by :user:`Guillaume Lemaitre <glemaitre>`.
+
+:mod:`sklearn.mixture`
+......................
+
+- |Fix| The initialization of :class:`mixture.GaussianMixture` from user-provided
+  `precisions_init` for `covariance_type` of `full` or `tied` was not correct,
+  and has been fixed.
+  :pr:`26416` by :user:`Yang Tao <mchikyt3>`.
+
 :mod:`sklearn.neighbors`
 ........................
 
@@ -58,12 +107,20 @@ Changelog
   when the input to the `param_distributions` parameter is a list of dicts.
   :pr:`26893` by :user:`Stefanie Senger <StefanieSenger>`.
 
+- |Fix| Neighbors based estimators now correctly work when `metric="minkowski"` and the
+  metric parameter `p` is in the range `0 < p < 1`, regardless of the `dtype` of `X`.
+  :pr:`26760` by :user:`Shreesha Kumar Bhat <Shreesha3112>`.
+
 :mod:`sklearn.preprocessing`
 ............................
 
 - |Fix| :class:`preprocessing.LabelEncoder` correctly accepts `y` as a keyword
   argument. :pr:`26940` by `Thomas Fan`_.
 
+- |Fix| :class:`preprocessing.OneHotEncoder` shows a more informative error message
+  when `sparse_output=True` and the output is configured to be pandas.
+  :pr:`26931` by `Thomas Fan`_.
+
 :mod:`sklearn.tree`
 ...................