Merge pull request scikit-learn#4010 from MechCoder/sparse_bug

[MRG] FIX: Bug in sparse coordinate solver in lazy centering

Author: agramfort

doc/whats_new.rst

@@ -227,6 +227,10 @@ Bug fixes
       :class:`cluster.AgglomerativeClustering` (and friends).
       This has been fixed By `Manoj Kumar`_
 
+    - Fix lazy centering of data in :func:`linear_model.enet_path` and
+      :func:`linear_model.lasso_path`. It was centered around one. It has
+      been changed to be centred around the origin. By `Manoj Kumar`_
+
 API changes summary
 -------------------
 

sklearn/linear_model/coordinate_descent.py

@@ -451,7 +451,7 @@ def enet_path(X, y, l1_ratio=0.5, eps=1e-3, n_alphas=100, alphas=None,
             # to be passed to the CD solver.
             X_sparse_scaling = params['X_mean'] / params['X_std']
         else:
-            X_sparse_scaling = np.ones(n_features)
+            X_sparse_scaling = np.zeros(n_features)
 
     X, y, X_mean, y_mean, X_std, precompute, Xy = \
         _pre_fit(X, y, Xy, precompute, normalize, fit_intercept, copy=False)

sklearn/linear_model/tests/test_coordinate_descent.py

@@ -585,6 +585,18 @@ def test_enet_path_positive():
         assert_true(np.all(pos_path_coef >= 0))
 
 
+def test_sparse_dense_descent_paths():
+    """
+    Test that dense and sparse input give the same input for descent paths.
+    """
+    X, y, _, _ = build_dataset(n_samples=50, n_features=20)
+    csr = sparse.csr_matrix(X)
+    for path in [enet_path, lasso_path]:
+        _, coefs, _ = path(X, y, fit_intercept=False)
+        _, sparse_coefs, _ = path(csr, y, fit_intercept=False)
+        assert_array_almost_equal(coefs, sparse_coefs)
+
+
 if __name__ == '__main__':
     import nose
     nose.runmodule()