scikit-learn · isaacknjama · Feb 10, 2021 · May 5, 2021 · May 5, 2021
diff --git a/sklearn/multiclass.py b/sklearn/multiclass.py
@@ -726,7 +726,8 @@ def predict(self, X):
            return self.classes_[(Y > 0).astype(int)]
        return self.classes_[Y.argmax(axis=1)]

-    def decision_function(self, X):
+    @_deprecate_positional_args
+    def decision_function(self, X, random_state='warn'):
        """Decision function for the OneVsOneClassifier.

        The decision values for the samples are computed by adding the
@@ -737,6 +738,8 @@ def decision_function(self, X):
        Parameters
        ----------
        X : array-like of shape (n_samples, n_features)
+        init: issues a warning by default and makes it possible to choose
+              between the past or future behaviour.

        Returns
        -------
@@ -765,9 +768,19 @@ def decision_function(self, X):
            return Y[:, 1]
        return Y

-    @property
-    def n_classes_(self):
-        return len(self.classes_)
+        if random_state == 'warn':
+            warnings.warn(
+                "'random_state' will be set to None starting from 1.2 which"
+                "means that results will differ at every function call."
+                "Set 'random_state' to an integer value to silence this"
+                "warning, or to 0 to keep the behavior of versions <1.0",
+                FutureWarning
+            )
+        random_state = 0
+
+        @property
+        def n_classes_(self):
+            return len(self.classes_)

    # TODO: Remove in 1.1
    # mypy error: Decorated property not supported

diff --git a/sklearn/svm/_base.py b/sklearn/svm/_base.py
@@ -560,7 +560,8 @@ def _validate_targets(self, y):

        return np.asarray(y, dtype=np.float64, order='C')

-    def decision_function(self, X):
+    @_deprecate_positional_args
+    def decision_function(self, X, random_state='warn'):
        """Evaluates the decision function for the samples in X.

        Parameters
@@ -591,15 +592,26 @@ def decision_function(self, X):
            return _ovr_decision_function(dec < 0, -dec, len(self.classes_))
        return dec

+        if random_state == 'warn':
+            warnings.warn(
+                "'random_state' will be set to None starting from 1.2 which"
+                "means that results will differ at every function call."
+                "Set 'random_state' to an integer value to silence this"
+                "warning, or to 0 to keep the behavior of versions <1.0",
+                FutureWarning
+            )
+        random_state = 0
+
    def predict(self, X):
        """Perform classification on samples in X.

        For an one-class model, +1 or -1 is returned.

        Parameters
        ----------
-        X : {array-like, sparse matrix} of shape (n_samples, n_features) or \
-                (n_samples_test, n_samples_train)
+        X : {array-like, sparse matrix} of shape
+            (n_samples, n_features) or \
+            (n_samples_test, n_samples_train)
            For kernel="precomputed", the expected shape of X is
            (n_samples_test, n_samples_train).


diff --git a/sklearn/utils/multiclass.py b/sklearn/utils/multiclass.py
@@ -461,4 +461,4 @@ def _ovr_decision_function(predictions, confidences, n_classes):
    # of 1 vote.
    transformed_confidences = (sum_of_confidences /
                               (3 * (np.abs(sum_of_confidences) + 1)))
-    return votes + transformed_confidences
+    return votes + transformed_confidences - (n_classes - 1) / 2