DOC Add link to Early Stopping example in Gradient Boosting (scikit-learn#27025)

Co-authored-by: Adrin Jalali <adrin.jalali@gmail.com>

Author: greyisbetter

doc/modules/ensemble.rst

@@ -740,8 +740,10 @@ of ``learning_rate`` require larger numbers of weak learners to maintain
 a constant training error. Empirical evidence suggests that small
 values of ``learning_rate`` favor better test error. [HTF]_
 recommend to set the learning rate to a small constant
-(e.g. ``learning_rate <= 0.1``) and choose ``n_estimators`` by early
-stopping. For a more detailed discussion of the interaction between
+(e.g. ``learning_rate <= 0.1``) and choose ``n_estimators`` large enough
+that early stopping applies,
+see :ref:`sphx_glr_auto_examples_ensemble_plot_gradient_boosting_early_stopping.py`
+for a more detailed discussion of the interaction between
 ``learning_rate`` and ``n_estimators`` see [R2007]_.
 
 Subsampling

examples/ensemble/plot_gradient_boosting_early_stopping.py

@@ -1,6 +1,6 @@
 """
 ===================================
-Early stopping of Gradient Boosting
+Early stopping in Gradient Boosting
 ===================================
 
 Gradient boosting is an ensembling technique where several weak learners

sklearn/ensemble/_gb.py

@@ -1282,6 +1282,8 @@ class GradientBoostingClassifier(ClassifierMixin, BaseGradientBoosting):
         improving in all of the previous ``n_iter_no_change`` numbers of
         iterations. The split is stratified.
         Values must be in the range `[1, inf)`.
+        See
+        :ref:`sphx_glr_auto_examples_ensemble_plot_gradient_boosting_early_stopping.py`.
 
         .. versionadded:: 0.20
 
@@ -1891,6 +1893,8 @@ class GradientBoostingRegressor(RegressorMixin, BaseGradientBoosting):
         improving in all of the previous ``n_iter_no_change`` numbers of
         iterations.
         Values must be in the range `[1, inf)`.
+        See
+        :ref:`sphx_glr_auto_examples_ensemble_plot_gradient_boosting_early_stopping.py`.
 
         .. versionadded:: 0.20
 

sklearn/linear_model/_passive_aggressive.py

@@ -35,11 +35,11 @@ class PassiveAggressiveClassifier(BaseSGDClassifier):
         .. versionadded:: 0.19
 
     early_stopping : bool, default=False
-        Whether to use early stopping to terminate training when validation.
+        Whether to use early stopping to terminate training when validation
         score is not improving. If set to True, it will automatically set aside
         a stratified fraction of training data as validation and terminate
-        training when validation score is not improving by at least tol for
-        n_iter_no_change consecutive epochs.
+        training when validation score is not improving by at least `tol` for
+        `n_iter_no_change` consecutive epochs.
 
         .. versionadded:: 0.20
 

sklearn/linear_model/_perceptron.py

@@ -68,11 +68,11 @@ class Perceptron(BaseSGDClassifier):
         See :term:`Glossary <random_state>`.
 
     early_stopping : bool, default=False
-        Whether to use early stopping to terminate training when validation.
+        Whether to use early stopping to terminate training when validation
         score is not improving. If set to True, it will automatically set aside
         a stratified fraction of training data as validation and terminate
-        training when validation score is not improving by at least tol for
-        n_iter_no_change consecutive epochs.
+        training when validation score is not improving by at least `tol` for
+        `n_iter_no_change` consecutive epochs.
 
         .. versionadded:: 0.20
 

sklearn/neural_network/_multilayer_perceptron.py

@@ -887,7 +887,7 @@ class MLPClassifier(ClassifierMixin, BaseMultilayerPerceptron):
         Whether to use early stopping to terminate training when validation
         score is not improving. If set to true, it will automatically set
         aside 10% of training data as validation and terminate training when
-        validation score is not improving by at least tol for
+        validation score is not improving by at least ``tol`` for
         ``n_iter_no_change`` consecutive epochs. The split is stratified,
         except in a multilabel setting.
         If early stopping is False, then the training stops when the training