FIX temporary fix for scikit-learn#4710

TomDLT · TomDLT · commit e1f9625653d6 · 2015-06-08T18:20:19.000+02:00
diff --git a/sklearn/linear_model/ridge.py b/sklearn/linear_model/ridge.py
@@ -313,11 +313,8 @@ def ridge_regression(X, y, alpha, sample_weight=None, solver='auto',
             raise ValueError("Sample weights must be 1D array or scalar")
 
         if solver != 'sag':
-            # Sample weight can be implemented via a simple rescaling.
-            # FIX: In the sparse case, with fit_intercept=True, we cannot
-            # center the data, so a rescaling is incorrect! See #4710.
-            # Sag solver can actually handle sample weights without rescaling,
-            # and can be used as a reference to solve #4710.
+            # SAG supports sample_weight directly. For other solvers,
+            # we implement sample_weight via a simple rescaling.
             X, y = _rescale_data(X, y, sample_weight)
 
     # There should be either 1 or n_targets penalties
@@ -405,6 +402,13 @@ def fit(self, X, y, sample_weight=None):
             X, y, self.fit_intercept, self.normalize, self.copy_X,
             sample_weight=sample_weight)
 
+        if sparse.issparse(X) and self.fit_intercept:
+            if self.solver == 'auto':
+                self.solver = 'sag'
+            elif self.solver != 'sag':
+                raise ValueError("Only sag solver can handle sparse data with"
+                                 " fit_intercept=True (temp fix for #4710)")
+
         self.coef_ = ridge_regression(X, y,
                                       alpha=self.alpha,
                                       sample_weight=sample_weight,
