[MRG+1] fix for erroneous max_iter and tol warnings for SGDClassifier when using partial_fit (#10053)

siftikha · amueller · commit f485a9edf92e · 2017-11-15T13:38:59.000-08:00
* partial fit warnings disabled

* partial fit warnings disabled for regressor

* style improved

* tests added

* pycodestyle passing

* rejiggered format

* fixed style issues
diff --git a/sklearn/linear_model/stochastic_gradient.py b/sklearn/linear_model/stochastic_gradient.py
@@ -82,7 +82,7 @@ def set_params(self, *args, **kwargs):
     def fit(self, X, y):
         """Fit model."""
 
-    def _validate_params(self, set_max_iter=True):
+    def _validate_params(self, set_max_iter=True, for_partial_fit=False):
         """Validate input params. """
         if not isinstance(self.shuffle, bool):
             raise ValueError("shuffle must be either True or False")
@@ -120,14 +120,15 @@ def _validate_params(self, set_max_iter=True):
             self._tol = None
 
         elif self.tol is None and self.max_iter is None:
-            warnings.warn(
-                "max_iter and tol parameters have been added in %s in 0.19. If"
-                " both are left unset, they default to max_iter=5 and tol=None"
-                ". If tol is not None, max_iter defaults to max_iter=1000. "
-                "From 0.21, default max_iter will be 1000, "
-                "and default tol will be 1e-3." % type(self).__name__,
-                FutureWarning)
-            # Before 0.19, default was n_iter=5
+            if not for_partial_fit:
+                warnings.warn(
+                    "max_iter and tol parameters have been "
+                    "added in %s in 0.19. If both are left unset, "
+                    "they default to max_iter=5 and tol=None. "
+                    "If tol is not None, max_iter defaults to max_iter=1000. "
+                    "From 0.21, default max_iter will be 1000, and"
+                    " default tol will be 1e-3." % type(self), FutureWarning)
+                # Before 0.19, default was n_iter=5
             max_iter = 5
         else:
             max_iter = self.max_iter if self.max_iter is not None else 1000
@@ -539,7 +540,7 @@ def partial_fit(self, X, y, classes=None, sample_weight=None):
         -------
         self : returns an instance of self.
         """
-        self._validate_params()
+        self._validate_params(for_partial_fit=True)
         if self.class_weight in ['balanced']:
             raise ValueError("class_weight '{0}' is not supported for "
                              "partial_fit. In order to use 'balanced' weights,"
@@ -984,7 +985,7 @@ def partial_fit(self, X, y, sample_weight=None):
         -------
         self : returns an instance of self.
         """
-        self._validate_params()
+        self._validate_params(for_partial_fit=True)
         return self._partial_fit(X, y, self.alpha, C=1.0,
                                  loss=self.loss,
                                  learning_rate=self.learning_rate, max_iter=1,
diff --git a/sklearn/linear_model/tests/test_sgd.py b/sklearn/linear_model/tests/test_sgd.py
@@ -1194,9 +1194,9 @@ def test_tol_parameter():
 def test_future_and_deprecation_warnings():
     # Test that warnings are raised. Will be removed in 0.21
 
-    def init(max_iter=None, tol=None, n_iter=None):
+    def init(max_iter=None, tol=None, n_iter=None, for_partial_fit=False):
         sgd = SGDClassifier(max_iter=max_iter, tol=tol, n_iter=n_iter)
-        sgd._validate_params()
+        sgd._validate_params(for_partial_fit=for_partial_fit)
 
     # When all default values are used
     msg_future = "max_iter and tol parameters have been added in "
@@ -1211,6 +1211,9 @@ def init(max_iter=None, tol=None, n_iter=None):
     assert_no_warnings(init, None, 1e-3, None)
     assert_no_warnings(init, 100, 1e-3, None)
 
+    # Test that for_partial_fit will not throw warnings for max_iter or tol
+    assert_no_warnings(init, None, None, None, True)
+
 
 @ignore_warnings(category=(DeprecationWarning, FutureWarning))
 def test_tol_and_max_iter_default_values():
