FIX : get rid of coef_init in Lasso in dict learning

agramfort · agramfort · commit faddb4565e07 · 2013-09-16T21:34:26.000+02:00
diff --git a/sklearn/decomposition/dict_learning.py b/sklearn/decomposition/dict_learning.py
@@ -106,8 +106,9 @@ def _sparse_encode(X, dictionary, gram, cov=None, algorithm='lasso_lars',
     elif algorithm == 'lasso_cd':
         alpha = float(regularization) / n_features  # account for scaling
         clf = Lasso(alpha=alpha, fit_intercept=False, precompute=gram,
-                    max_iter=max_iter)
-        clf.fit(dictionary.T, X.T, coef_init=init)
+                    max_iter=max_iter, warm_start=True)
+        clf.coef_ = init
+        clf.fit(dictionary.T, X.T)
         new_code = clf.coef_
 
     elif algorithm == 'lars':
diff --git a/sklearn/linear_model/coordinate_descent.py b/sklearn/linear_model/coordinate_descent.py
@@ -552,7 +552,7 @@ def __init__(self, alpha=1.0, l1_ratio=0.5, fit_intercept=True,
         self.positive = positive
         self.intercept_ = 0.0
 
-    def fit(self, X, y, coef_init=None):
+    def fit(self, X, y):
         """Fit model with coordinate descent.
 
         Parameters
@@ -563,10 +563,6 @@ def fit(self, X, y, coef_init=None):
         y : ndarray, shape = (n_samples,) or (n_samples, n_targets)
             Target
 
-        coef_init : ndarray of shape n_features or (n_targets, n_features)
-            The initial coefficients to warm-start the optimization
-            WARNING : ignored and will be deprecated in 0.15
-
         Notes
         -----
 
