scikit-learn · glemaitre · May 11, 2022 · May 7, 2022 · May 7, 2022 · May 7, 2022
diff --git a/doc/whats_new/v1.2.rst b/doc/whats_new/v1.2.rst
@@ -58,6 +58,13 @@ Changelog
   :pr:`10468` by :user:`Ruben <icfly2>` and :pr:`22993` by
   :user:`Jovan Stojanovic <jovan-stojanovic>`.
 
+:mod:`sklearn.feature_selection`
+................................
+
+- |Fix| The `partial_fit` method of :class:`feature_selection.SelectFromModel`
+  now conducts validation for `max_features` and `feature_names_in` parameters.
+  :pr:`23299` by :user:`Long Bao <lorentzbao>`.
+
 Code and Documentation Contributors
 -----------------------------------
 

diff --git a/sklearn/feature_selection/_from_model.py b/sklearn/feature_selection/_from_model.py
@@ -10,11 +10,10 @@
 from ._base import _get_feature_importances
 from ..base import BaseEstimator, clone, MetaEstimatorMixin
 from ..utils._tags import _safe_tags
-from ..utils.validation import check_is_fitted
+from ..utils.validation import check_is_fitted, check_scalar, _num_features
 
 from ..exceptions import NotFittedError
 from ..utils.metaestimators import available_if
-from ..utils.validation import check_scalar
 
 
 def _calculate_threshold(estimator, importances, threshold):
@@ -287,34 +286,17 @@ def _get_support_mask(self):
         mask[scores < threshold] = False
         return mask
 
-    def fit(self, X, y=None, **fit_params):
-        """Fit the SelectFromModel meta-transformer.
-
-        Parameters
-        ----------
-        X : array-like of shape (n_samples, n_features)
-            The training input samples.
-
-        y : array-like of shape (n_samples,), default=None
-            The target values (integers that correspond to classes in
-            classification, real numbers in regression).
-
-        **fit_params : dict
-            Other estimator specific parameters.
-
-        Returns
-        -------
-        self : object
-            Fitted estimator.
-        """
+    def _check_max_features(self, X):
         if self.max_features is not None:
+            n_features = _num_features(X)
+
             if isinstance(self.max_features, numbers.Integral):
                 check_scalar(
                     self.max_features,
                     "max_features",
                     numbers.Integral,
                     min_val=0,
-                    max_val=len(X[0]),
+                    max_val=n_features,
                 )
                 self.max_features_ = self.max_features
             elif callable(self.max_features):
@@ -324,7 +306,7 @@ def fit(self, X, y=None, **fit_params):
                     "max_features(X)",
                     numbers.Integral,
                     min_val=0,
-                    max_val=len(X[0]),
+                    max_val=n_features,
                 )
                 self.max_features_ = max_features
             else:
@@ -333,6 +315,28 @@ def fit(self, X, y=None, **fit_params):
                     f" 'X' as input. Got {self.max_features} instead."
                 )
 
+    def fit(self, X, y=None, **fit_params):
+        """Fit the SelectFromModel meta-transformer.
+
+        Parameters
+        ----------
+        X : array-like of shape (n_samples, n_features)
+            The training input samples.
+
+        y : array-like of shape (n_samples,), default=None
+            The target values (integers that correspond to classes in
+            classification, real numbers in regression).
+
+        **fit_params : dict
+            Other estimator specific parameters.
+
+        Returns
+        -------
+        self : object
+            Fitted estimator.
+        """
+        self._check_max_features(X)
+
         if self.prefit:
             try:
                 check_is_fitted(self.estimator)
@@ -385,6 +389,8 @@ def partial_fit(self, X, y=None, **fit_params):
         self : object
             Fitted estimator.
         """
+        self._check_max_features(X)
+
         if self.prefit:
             if not hasattr(self, "estimator_"):
                 try:
@@ -397,9 +403,16 @@ def partial_fit(self, X, y=None, **fit_params):
                 self.estimator_ = deepcopy(self.estimator)
             return self
 
-        if not hasattr(self, "estimator_"):
+        first_call = not hasattr(self, "estimator_")
+        if first_call:
             self.estimator_ = clone(self.estimator)
         self.estimator_.partial_fit(X, y, **fit_params)
+
+        if hasattr(self.estimator_, "feature_names_in_"):
+            self.feature_names_in_ = self.estimator_.feature_names_in_
+        else:
+            self._check_feature_names(X, reset=first_call)
+
         return self
 
     @property

diff --git a/sklearn/feature_selection/tests/test_from_model.py b/sklearn/feature_selection/tests/test_from_model.py
@@ -607,3 +607,40 @@ def importance_getter(estimator):
         warnings.simplefilter("error", UserWarning)
 
         selector.transform(X.iloc[1:3])
+
+
+@pytest.mark.parametrize(
+    "error, err_msg, max_features",
+    (
+        [ValueError, "max_features == 10, must be <= 4", 10],
+        [TypeError, "'max_features' must be either an int or a callable", "a"],
+        [ValueError, r"max_features\(X\) == 5, must be <= 4", lambda x: x.shape[1] + 1],
+    ),
+)
+def test_partial_fit_validate_max_features(error, err_msg, max_features):
+    """Test that partial_fit from SelectFromModel validates `max_features`."""
+    X, y = datasets.make_classification(
+        n_samples=100,
+        n_features=4,
+        random_state=0,
+    )
+
+    with pytest.raises(error, match=err_msg):
+        SelectFromModel(
+            estimator=SGDClassifier(), max_features=max_features
+        ).partial_fit(X, y, classes=[0, 1])
+
+
+@pytest.mark.parametrize("as_frame", [True, False])
+def test_partial_fit_validate_feature_names(as_frame):
+    """Test that partial_fit from SelectFromModel validates `feature_names_in_`."""
+    pytest.importorskip("pandas")
+    X, y = datasets.load_iris(as_frame=as_frame, return_X_y=True)
+
+    selector = SelectFromModel(estimator=SGDClassifier(), max_features=4).partial_fit(
+        X, y, classes=[0, 1, 2]
+    )
+    if as_frame:
+        assert_array_equal(selector.feature_names_in_, X.columns)
+    else:
+        assert not hasattr(selector, "feature_names_in_")
diff --git a/sklearn/utils/estimator_checks.py b/sklearn/utils/estimator_checks.py
@@ -31,6 +31,7 @@
 from ..linear_model import LogisticRegression
 from ..linear_model import RANSACRegressor
 from ..linear_model import Ridge
+from ..linear_model import SGDRegressor
 
 from ..base import (
     clone,
@@ -44,6 +45,7 @@
 from ..metrics import accuracy_score, adjusted_rand_score, f1_score
 from ..random_projection import BaseRandomProjection
 from ..feature_selection import SelectKBest
+from ..feature_selection import SelectFromModel
 from ..pipeline import make_pipeline
 from ..exceptions import DataConversionWarning
 from ..exceptions import NotFittedError
@@ -389,6 +391,9 @@ def _construct_instance(Estimator):
                 estimator = Estimator(LinearRegression())
             elif issubclass(Estimator, RegressorMixin):
                 estimator = Estimator(Ridge())
+            elif issubclass(Estimator, SelectFromModel):
+                # Increases coverage because SGDRegressor has partial_fit
+                estimator = Estimator(SGDRegressor(random_state=0))
             else:
                 estimator = Estimator(LogisticRegression(C=1))
         elif required_parameters in (["estimators"],):