DEP change the default of gamma in SVM (scikit-learn#13801)

Author: glemaitre

doc/modules/compose.rst

@@ -216,12 +216,12 @@ object::
      >>> from sklearn.datasets import load_digits
      >>> digits = load_digits()
      >>> pca1 = PCA()
-     >>> svm1 = SVC(gamma='scale')
+     >>> svm1 = SVC()
      >>> pipe = Pipeline([('reduce_dim', pca1), ('clf', svm1)])
      >>> pipe.fit(digits.data, digits.target)
      ... # doctest: +NORMALIZE_WHITESPACE, +ELLIPSIS
      Pipeline(memory=None,
-              steps=[('reduce_dim', PCA(...)), ('clf', SVC(...))], 
+              steps=[('reduce_dim', PCA(...)), ('clf', SVC(...))],
               verbose=False)
      >>> # The pca instance can be inspected directly
      >>> print(pca1.components_) # doctest: +NORMALIZE_WHITESPACE, +ELLIPSIS
@@ -238,7 +238,7 @@ object::
 
      >>> cachedir = mkdtemp()
      >>> pca2 = PCA()
-     >>> svm2 = SVC(gamma='scale')
+     >>> svm2 = SVC()
      >>> cached_pipe = Pipeline([('reduce_dim', pca2), ('clf', svm2)],
      ...                        memory=cachedir)
      >>> cached_pipe.fit(digits.data, digits.target)

doc/modules/ensemble.rst

@@ -947,7 +947,7 @@ Vector Machine, a Decision Tree, and a K-nearest neighbor classifier::
    >>> # Training classifiers
    >>> clf1 = DecisionTreeClassifier(max_depth=4)
    >>> clf2 = KNeighborsClassifier(n_neighbors=7)
-   >>> clf3 = SVC(gamma='scale', kernel='rbf', probability=True)
+   >>> clf3 = SVC(kernel='rbf', probability=True)
    >>> eclf = VotingClassifier(estimators=[('dt', clf1), ('knn', clf2), ('svc', clf3)],
    ...                         voting='soft', weights=[2, 1, 2])
 

doc/modules/model_evaluation.rst

@@ -100,7 +100,7 @@ Usage examples:
     >>> from sklearn.model_selection import cross_val_score
     >>> iris = datasets.load_iris()
     >>> X, y = iris.data, iris.target
-    >>> clf = svm.SVC(gamma='scale', random_state=0)
+    >>> clf = svm.SVC(random_state=0)
     >>> cross_val_score(clf, X, y, scoring='recall_macro',
     ...                 cv=5)  # doctest: +ELLIPSIS, +NORMALIZE_WHITESPACE
     array([0.96..., 0.96..., 0.96..., 0.93..., 1.        ])
@@ -1964,7 +1964,7 @@ Next, let's compare the accuracy of ``SVC`` and ``most_frequent``::
 We see that ``SVC`` doesn't do much better than a dummy classifier. Now, let's
 change the kernel::
 
-  >>> clf = SVC(gamma='scale', kernel='rbf', C=1).fit(X_train, y_train)
+  >>> clf = SVC(kernel='rbf', C=1).fit(X_train, y_train)
   >>> clf.score(X_test, y_test)  # doctest: +ELLIPSIS
   0.94...
 

doc/modules/model_persistence.rst

@@ -23,7 +23,7 @@ persistence model, namely `pickle <https://docs.python.org/2/library/pickle.html
 
   >>> from sklearn import svm
   >>> from sklearn import datasets
-  >>> clf = svm.SVC(gamma='scale')
+  >>> clf = svm.SVC()
   >>> iris = datasets.load_iris()
   >>> X, y = iris.data, iris.target
   >>> clf.fit(X, y)  # doctest: +NORMALIZE_WHITESPACE

doc/modules/svm.rst

@@ -75,7 +75,7 @@ n_features]`` holding the training samples, and an array y of class labels
     >>> from sklearn import svm
     >>> X = [[0, 0], [1, 1]]
     >>> y = [0, 1]
-    >>> clf = svm.SVC(gamma='scale')
+    >>> clf = svm.SVC()
     >>> clf.fit(X, y)  # doctest: +NORMALIZE_WHITESPACE
     SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,
         decision_function_shape='ovr', degree=3, gamma='scale', kernel='rbf',
@@ -119,7 +119,7 @@ n_classes)``.
 
     >>> X = [[0], [1], [2], [3]]
     >>> Y = [0, 1, 2, 3]
-    >>> clf = svm.SVC(gamma='scale', decision_function_shape='ovo')
+    >>> clf = svm.SVC(decision_function_shape='ovo')
     >>> clf.fit(X, Y) # doctest: +NORMALIZE_WHITESPACE
     SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,
         decision_function_shape='ovo', degree=3, gamma='scale', kernel='rbf',
@@ -314,8 +314,8 @@ Vector Regression depends only on a subset of the training data,
 because the cost function for building the model ignores any training
 data close to the model prediction.
 
-There are three different implementations of Support Vector Regression: 
-:class:`SVR`, :class:`NuSVR` and :class:`LinearSVR`. :class:`LinearSVR` 
+There are three different implementations of Support Vector Regression:
+:class:`SVR`, :class:`NuSVR` and :class:`LinearSVR`. :class:`LinearSVR`
 provides a faster implementation than :class:`SVR` but only considers
 linear kernels, while :class:`NuSVR` implements a slightly different
 formulation than :class:`SVR` and :class:`LinearSVR`. See
@@ -331,7 +331,7 @@ floating point values instead of integer values::
     >>> clf = svm.SVR()
     >>> clf.fit(X, y) # doctest: +NORMALIZE_WHITESPACE
     SVR(C=1.0, cache_size=200, coef0=0.0, degree=3, epsilon=0.1,
-        gamma='auto_deprecated', kernel='rbf', max_iter=-1, shrinking=True,
+        gamma='scale', kernel='rbf', max_iter=-1, shrinking=True,
         tol=0.001, verbose=False)
     >>> clf.predict([[1, 1]])
     array([1.5])
@@ -349,7 +349,7 @@ Density estimation, novelty detection
 =======================================
 
 The class :class:`OneClassSVM` implements a One-Class SVM which is used in
-outlier detection. 
+outlier detection.
 
 See :ref:`outlier_detection` for the description and usage of OneClassSVM.
 
@@ -542,7 +542,7 @@ test vectors must be provided.
     >>> gram = np.dot(X, X.T)
     >>> clf.fit(gram, y) # doctest: +NORMALIZE_WHITESPACE
     SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,
-        decision_function_shape='ovr', degree=3, gamma='auto_deprecated',
+        decision_function_shape='ovr', degree=3, gamma='scale',
         kernel='precomputed', max_iter=-1,
         probability=False, random_state=None, shrinking=True, tol=0.001,
         verbose=False)

doc/tutorial/basic/tutorial.rst

@@ -216,7 +216,7 @@ persistence model, `pickle <https://docs.python.org/2/library/pickle.html>`_::
 
   >>> from sklearn import svm
   >>> from sklearn import datasets
-  >>> clf = svm.SVC(gamma='scale')
+  >>> clf = svm.SVC()
   >>> iris = datasets.load_iris()
   >>> X, y = iris.data, iris.target
   >>> clf.fit(X, y)  # doctest: +NORMALIZE_WHITESPACE
@@ -291,7 +291,7 @@ maintained::
     >>> from sklearn import datasets
     >>> from sklearn.svm import SVC
     >>> iris = datasets.load_iris()
-    >>> clf = SVC(gamma='scale')
+    >>> clf = SVC()
     >>> clf.fit(iris.data, iris.target)  # doctest: +NORMALIZE_WHITESPACE
     SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,
       decision_function_shape='ovr', degree=3, gamma='scale', kernel='rbf',
@@ -329,14 +329,14 @@ once will overwrite what was learned by any previous ``fit()``::
   >>> clf = SVC()
   >>> clf.set_params(kernel='linear').fit(X, y)  # doctest: +NORMALIZE_WHITESPACE
   SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,
-    decision_function_shape='ovr', degree=3, gamma='auto_deprecated',
+    decision_function_shape='ovr', degree=3, gamma='scale',
     kernel='linear', max_iter=-1,
     probability=False, random_state=None, shrinking=True, tol=0.001,
     verbose=False)
   >>> clf.predict(X[:5])
   array([0, 0, 0, 0, 0])
 
-  >>> clf.set_params(kernel='rbf', gamma='scale').fit(X, y)  # doctest: +NORMALIZE_WHITESPACE
+  >>> clf.set_params(kernel='rbf').fit(X, y)  # doctest: +NORMALIZE_WHITESPACE
   SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,
     decision_function_shape='ovr', degree=3, gamma='scale', kernel='rbf',
     max_iter=-1, probability=False,
@@ -363,8 +363,7 @@ the target data fit upon::
     >>> X = [[1, 2], [2, 4], [4, 5], [3, 2], [3, 1]]
     >>> y = [0, 0, 1, 1, 2]
 
-    >>> classif = OneVsRestClassifier(estimator=SVC(gamma='scale',
-    ...                                             random_state=0))
+    >>> classif = OneVsRestClassifier(estimator=SVC(random_state=0))
     >>> classif.fit(X, y).predict(X)
     array([0, 0, 1, 1, 2])
 

doc/tutorial/statistical_inference/supervised_learning.rst

@@ -463,7 +463,7 @@ classification --:class:`SVC` (Support Vector Classification).
     >>> svc = svm.SVC(kernel='linear')
     >>> svc.fit(iris_X_train, iris_y_train)    # doctest: +NORMALIZE_WHITESPACE
     SVC(C=1.0, break_ties=False, cache_size=200, class_weight=None, coef0=0.0,
-        decision_function_shape='ovr', degree=3, gamma='auto_deprecated',
+        decision_function_shape='ovr', degree=3, gamma='scale',
         kernel='linear', max_iter=-1,
         probability=False, random_state=None, shrinking=True, tol=0.001,
         verbose=False)

examples/preprocessing/plot_discretization_classification.py

@@ -83,7 +83,7 @@ def get_name(estimator):
     (GradientBoostingClassifier(n_estimators=50, random_state=0), {
         'learning_rate': np.logspace(-4, 0, 10)
     }),
-    (SVC(random_state=0, gamma='scale'), {
+    (SVC(random_state=0), {
         'C': np.logspace(-2, 7, 10)
     }),
 ]

sklearn/ensemble/tests/test_bagging.py

@@ -71,7 +71,7 @@ def test_classification():
                            Perceptron(tol=1e-3),
                            DecisionTreeClassifier(),
                            KNeighborsClassifier(),
-                           SVC(gamma="scale")]:
+                           SVC()]:
         for params in grid:
             BaggingClassifier(base_estimator=base_estimator,
                               random_state=rng,
@@ -117,17 +117,15 @@ def fit(self, X, y):
             for f in ['predict', 'predict_proba', 'predict_log_proba', 'decision_function']:
                 # Trained on sparse format
                 sparse_classifier = BaggingClassifier(
-                    base_estimator=CustomSVC(gamma='scale',
-                                             decision_function_shape='ovr'),
+                    base_estimator=CustomSVC(decision_function_shape='ovr'),
                     random_state=1,
                     **params
                 ).fit(X_train_sparse, y_train)
                 sparse_results = getattr(sparse_classifier, f)(X_test_sparse)
 
                 # Trained on dense format
                 dense_classifier = BaggingClassifier(
-                    base_estimator=CustomSVC(gamma='scale',
-                                             decision_function_shape='ovr'),
+                    base_estimator=CustomSVC(decision_function_shape='ovr'),
                     random_state=1,
                     **params
                 ).fit(X_train, y_train)
@@ -155,7 +153,7 @@ def test_regression():
                            DummyRegressor(),
                            DecisionTreeRegressor(),
                            KNeighborsRegressor(),
-                           SVR(gamma='scale')]:
+                           SVR()]:
         for params in grid:
             BaggingRegressor(base_estimator=base_estimator,
                              random_state=rng,
@@ -201,15 +199,15 @@ def fit(self, X, y):
 
             # Trained on sparse format
             sparse_classifier = BaggingRegressor(
-                base_estimator=CustomSVR(gamma='scale'),
+                base_estimator=CustomSVR(),
                 random_state=1,
                 **params
             ).fit(X_train_sparse, y_train)
             sparse_results = sparse_classifier.predict(X_test_sparse)
 
             # Trained on dense format
             dense_results = BaggingRegressor(
-                base_estimator=CustomSVR(gamma='scale'),
+                base_estimator=CustomSVR(),
                 random_state=1,
                 **params
             ).fit(X_train, y_train).predict(X_test)
@@ -334,7 +332,7 @@ def test_oob_score_classification():
                                                         iris.target,
                                                         random_state=rng)
 
-    for base_estimator in [DecisionTreeClassifier(), SVC(gamma="scale")]:
+    for base_estimator in [DecisionTreeClassifier(), SVC()]:
         clf = BaggingClassifier(base_estimator=base_estimator,
                                 n_estimators=100,
                                 bootstrap=True,
@@ -464,8 +462,7 @@ def test_parallel_classification():
     assert_array_almost_equal(y1, y3)
 
     # decision_function
-    ensemble = BaggingClassifier(SVC(gamma='scale',
-                                     decision_function_shape='ovr'),
+    ensemble = BaggingClassifier(SVC(decision_function_shape='ovr'),
                                  n_jobs=3,
                                  random_state=0).fit(X_train, y_train)
 
@@ -482,8 +479,7 @@ def test_parallel_classification():
                          "".format(X_test.shape[1], X_err.shape[1]),
                          ensemble.decision_function, X_err)
 
-    ensemble = BaggingClassifier(SVC(gamma='scale',
-                                     decision_function_shape='ovr'),
+    ensemble = BaggingClassifier(SVC(decision_function_shape='ovr'),
                                  n_jobs=1,
                                  random_state=0).fit(X_train, y_train)
 
@@ -529,7 +525,7 @@ def test_gridsearch():
     parameters = {'n_estimators': (1, 2),
                   'base_estimator__C': (1, 2)}
 
-    GridSearchCV(BaggingClassifier(SVC(gamma="scale")),
+    GridSearchCV(BaggingClassifier(SVC()),
                  parameters,
                  scoring="roc_auc").fit(X, y)
 
@@ -578,7 +574,7 @@ def test_base_estimator():
 
     assert isinstance(ensemble.base_estimator_, DecisionTreeRegressor)
 
-    ensemble = BaggingRegressor(SVR(gamma='scale'),
+    ensemble = BaggingRegressor(SVR(),
                                 n_jobs=3,
                                 random_state=0).fit(X_train, y_train)
     assert isinstance(ensemble.base_estimator_, SVR)

sklearn/ensemble/tests/test_voting.py

@@ -325,7 +325,7 @@ def test_sample_weight():
     """Tests sample_weight parameter of VotingClassifier"""
     clf1 = LogisticRegression(random_state=123)
     clf2 = RandomForestClassifier(random_state=123)
-    clf3 = SVC(gamma='scale', probability=True, random_state=123)
+    clf3 = SVC(probability=True, random_state=123)
     eclf1 = VotingClassifier(estimators=[
         ('lr', clf1), ('rf', clf2), ('svc', clf3)],
         voting='soft').fit(X, y, sample_weight=np.ones((len(y),)))