Merge branch 'main' into submodulev3

build_tools/azure/install.sh

@@ -139,7 +139,7 @@ scikit_learn_install() {
         python setup.py develop
     fi
 
-    ccache -s
+    ccache -s || echo "ccache not installed, skipping ccache statistics"
 }
 
 main() {

build_tools/azure/python_nogil_lock.txt

@@ -7,23 +7,21 @@
 --index-url https://d1yxz45j0ypngg.cloudfront.net/
 --extra-index-url https://pypi.org/simple
 
-contourpy==1.0.7
-    # via
-    #   -r /scikit-learn/build_tools/azure/python_nogil_requirements.txt
-    #   matplotlib
-cycler==0.11.0
+contourpy==1.1.1
+    # via matplotlib
+cycler==0.12.1
     # via matplotlib
-cython==0.29.33
+cython==3.0.8
     # via -r /scikit-learn/build_tools/azure/python_nogil_requirements.txt
-exceptiongroup==1.1.2
+exceptiongroup==1.2.0
     # via pytest
 execnet==2.0.2
     # via pytest-xdist
-fonttools==4.42.0
+fonttools==4.49.0
     # via matplotlib
 iniconfig==2.0.0
     # via pytest
-joblib==1.3.1
+joblib==1.3.2
     # via -r /scikit-learn/build_tools/azure/python_nogil_requirements.txt
 kiwisolver==1.4.4
     # via matplotlib
@@ -35,29 +33,29 @@ numpy==1.24.0
     #   contourpy
     #   matplotlib
     #   scipy
-packaging==23.1
+packaging==23.2
     # via
     #   matplotlib
     #   pytest
 pillow==9.5.0
     # via matplotlib
-pluggy==1.2.0
+pluggy==1.4.0
     # via pytest
 pyparsing==3.1.1
     # via matplotlib
-pytest==7.4.0
+pytest==8.0.1
     # via
     #   -r /scikit-learn/build_tools/azure/python_nogil_requirements.txt
     #   pytest-xdist
-pytest-xdist==3.3.1
+pytest-xdist==3.5.0
     # via -r /scikit-learn/build_tools/azure/python_nogil_requirements.txt
 python-dateutil==2.8.2
     # via matplotlib
 scipy==1.9.3
     # via -r /scikit-learn/build_tools/azure/python_nogil_requirements.txt
 six==1.16.0
     # via python-dateutil
-threadpoolctl==3.2.0
+threadpoolctl==3.3.0
     # via -r /scikit-learn/build_tools/azure/python_nogil_requirements.txt
 tomli==2.0.1
     # via pytest

build_tools/azure/python_nogil_requirements.txt

@@ -6,9 +6,6 @@
 # the latest cython will be picked up from PyPI, rather than the one from the
 # python-nogil index
 matplotlib
-# 2023-07-15 contourpy 1.1 needs meson which needs a recent pybind version
-# which is not available for python/nogil ...
-contourpy<1.1
 numpy
 scipy
 cython

build_tools/azure/windows.yml

@@ -37,7 +37,7 @@ jobs:
         architecture: 'x86'
       displayName: Use 32 bit System Python
       condition: and(succeeded(), eq(variables['PYTHON_ARCH'], '32'))
-    - bash: ./build_tools/azure/install_win.sh
+    - bash: ./build_tools/azure/install.sh
       displayName: 'Install'
     - bash: ./build_tools/azure/test_script.sh
       displayName: 'Test Library'

doc/modules/calibration.rst

@@ -74,10 +74,14 @@ by showing the number of samples in each predicted probability bin.
 
 .. currentmodule:: sklearn.linear_model
 
-:class:`LogisticRegression` returns well calibrated predictions by default as it has a
+:class:`LogisticRegression` is more likely to return well calibrated predictions by itself as it has a
 canonical link function for its loss, i.e. the logit-link for the :ref:`log_loss`.
-This leads to the so-called **balance property**, see [8]_ and
-:ref:`Logistic_regression`.
+In the unpenalized case, this leads to the so-called **balance property**, see [8]_ and :ref:`Logistic_regression`.
+In the plot above, data is generated according to a linear mechanism, which is
+consistent with the :class:`LogisticRegression` model (the model is 'well specified'),
+and the value of the regularization parameter `C` is tuned to be
+appropriate (neither too strong nor too low). As a consequence, this model returns
+accurate predictions from its `predict_proba` method.
 In contrast to that, the other shown models return biased probabilities; with
 different biases per model.
 

doc/modules/ensemble.rst

@@ -80,7 +80,8 @@ are not yet supported, for instance some loss functions.
 
 .. topic:: Examples:
 
- * :ref:`sphx_glr_auto_examples_inspection_plot_partial_dependence.py`
+  * :ref:`sphx_glr_auto_examples_inspection_plot_partial_dependence.py`
+  * :ref:`sphx_glr_auto_examples_ensemble_plot_forest_hist_grad_boosting_comparison.py`
 
 Usage
 ^^^^^
@@ -129,6 +130,8 @@ Note that for technical reasons, using a callable as a scorer is significantly s
 than using the loss. By default, early-stopping is performed if there are at least
 10,000 samples in the training set, using the validation loss.
 
+.. _nan_support_hgbt:
+
 Missing values support
 ^^^^^^^^^^^^^^^^^^^^^^
 
@@ -167,6 +170,10 @@ If no missing values were encountered for a given feature during training,
 then samples with missing values are mapped to whichever child has the most
 samples.
 
+.. topic:: Examples:
+
+  * :ref:`sphx_glr_auto_examples_ensemble_plot_hgbt_regression.py`
+
 .. _sw_hgbdt:
 
 Sample weight support
@@ -331,6 +338,7 @@ Also, monotonic constraints are not supported for multiclass classification.
 .. topic:: Examples:
 
   * :ref:`sphx_glr_auto_examples_ensemble_plot_monotonic_constraints.py`
+  * :ref:`sphx_glr_auto_examples_ensemble_plot_hgbt_regression.py`
 
 .. _interaction_cst_hgbt:
 

doc/modules/linear_model.rst

@@ -685,7 +685,7 @@ orthogonal matching pursuit can approximate the optimum solution vector with a
 fixed number of non-zero elements:
 
 .. math::
-    \underset{w}{\operatorname{arg\,min\,}}  ||y - Xw||_2^2 \text{ subject to } ||w||_0 \leq n_{\text{nonzero\_coefs}}
+    \underset{w}{\operatorname{arg\,min\,}}  ||y - Xw||_2^2 \text{ subject to } ||w||_0 \leq n_{\text{nonzero_coefs}}
 
 Alternatively, orthogonal matching pursuit can target a specific error instead
 of a specific number of non-zero coefficients. This can be expressed as:

doc/templates/index.html

@@ -70,8 +70,8 @@ <h4 class="sk-landing-subheader text-white font-italic mb-3">Machine Learning in
           and <a href="supervised_learning.html#supervised-learning">more...</a></p>
         </div>
         <div class="overflow-hidden mx-2 text-center flex-fill">
-          <a href="auto_examples/ensemble/plot_adaboost_regression.html"  aria-label="Regression">
-          <img src="_images/sphx_glr_plot_adaboost_regression_thumb.png" class="sk-index-img" alt="Decision Tree Regression with AdaBoost">
+          <a href="auto_examples/ensemble/plot_hgbt_regression.html"  aria-label="Regression">
+          <img src="_images/sphx_glr_plot_hgbt_regression_002.png" class="sk-index-img" alt="Decision Tree Regression with HGBT">
           </a>
         </div>
           <a href="auto_examples/index.html#examples" class="sk-btn-primary btn text-white btn-block" role="button">Examples</a>

doc/whats_new/v1.5.rst

@@ -123,11 +123,24 @@ Changelog
 :mod:`sklearn.linear_model`
 ...........................
 
+- |Fix| :class:`linear_model.ElasticNet`, :class:`linear_model.ElasticNetCV`,
+  :class:`linear_model.Lasso` and :class:`linear_model.LassoCV` now explicitly don't
+  accept large sparse data formats. :pr:`27576` by :user:`Stefanie Senger
+  <StefanieSenger>`.
+
 - |API| :class:`linear_model.RidgeCV` and :class:`linear_model.RidgeClassifierCV`
   will now allow `alpha=0` when `cv != None`, which is consistent with
   :class:`linear_model.Ridge` and :class:`linear_model.RidgeClassifier`.
   :pr:`28425` by :user:`Lucy Liu <lucyleeow>`.
 
+:mod:`sklearn.manifold`
+.......................
+
+- |API| Deprecates `n_iter` in favor of `max_iter` in :class:`manifold.TSNE`.
+  `n_iter` will be removed in version 1.7. This makes :class:`manifold.TSNE`
+  consistent with the rest of the estimators. :pr:`28471` by
+  :user:`Lucy Liu <lucyleeow>`
+
 :mod:`sklearn.metrics`
 ......................
 

examples/ensemble/plot_adaboost_regression.py

@@ -9,6 +9,10 @@
 regressor. As the number of boosts is increased the regressor can fit more
 detail.
 
+See :ref:`sphx_glr_auto_examples_ensemble_plot_hgbt_regression.py` for an
+example showcasing the benefits of using more efficient regression models such
+as :class:`~ensemble.HistGradientBoostingRegressor`.
+
 .. [1] `H. Drucker, "Improving Regressors using Boosting Techniques", 1997.
         <https://citeseerx.ist.psu.edu/doc_view/pid/8d49e2dedb817f2c3330e74b63c5fc86d2399ce3>`_
 

examples/ensemble/plot_forest_hist_grad_boosting_comparison.py

@@ -22,7 +22,9 @@
 the predicted value. RFs, on the other hand, are based on bagging and use a
 majority vote to predict the outcome.
 
-For more information on ensemble models, see the :ref:`User Guide <ensemble>`.
+See the :ref:`User Guide <ensemble>` for more information on ensemble models or
+see :ref:`sphx_glr_auto_examples_ensemble_plot_hgbt_regression.py` for an
+example showcasing some other features of HGBT models.
 """
 
 # Author:  Arturo Amor <david-arturo.amor-quiroz@inria.fr>

examples/ensemble/plot_gradient_boosting_categorical.py

@@ -21,6 +21,10 @@
 We will work with the Ames Iowa Housing dataset which consists of numerical
 and categorical features, where the houses' sales prices is the target.
 
+See :ref:`sphx_glr_auto_examples_ensemble_plot_hgbt_regression.py` for an
+example showcasing some other features of
+:class:`~ensemble.HistGradientBoostingRegressor`.
+
 """
 
 # %%

examples/ensemble/plot_gradient_boosting_quantile.py

@@ -4,7 +4,9 @@
 =====================================================
 
 This example shows how quantile regression can be used to create prediction
-intervals.
+intervals. See :ref:`sphx_glr_auto_examples_ensemble_plot_hgbt_regression.py`
+for an example showcasing some other features of
+:class:`~ensemble.HistGradientBoostingRegressor`.
 
 """
 

examples/ensemble/plot_gradient_boosting_regression.py

@@ -11,7 +11,10 @@
 and 500 regression trees of depth 4.
 
 Note: For larger datasets (n_samples >= 10000), please refer to
-:class:`~sklearn.ensemble.HistGradientBoostingRegressor`.
+:class:`~sklearn.ensemble.HistGradientBoostingRegressor`. See
+:ref:`sphx_glr_auto_examples_ensemble_plot_hgbt_regression.py` for an example
+showcasing some other advantages of
+:class:`~ensemble.HistGradientBoostingRegressor`.
 
 """