FIX OrthogonalMatchingPursuit normalized twice

Author: vene

sklearn/linear_model/omp.py

@@ -679,13 +679,6 @@ def fit(self, X, y, Gram=None, Xy=None):
                                             self.tol, norms_sq,
                                             copy_Gram, True).T
 
-        if self.normalize:
-            nonzeros = np.flatnonzero(X_std)
-            scaling = X_std[nonzeros]
-            if self.coef_.ndim == 2:
-                scaling = scaling[np.newaxis, :]
-            self.coef_[:, nonzeros] /= scaling
-
         self._set_intercept(X_mean, y_mean, X_std)
         return self
 

sklearn/linear_model/tests/test_omp.py

@@ -14,8 +14,10 @@
 
 from sklearn.linear_model import (orthogonal_mp, orthogonal_mp_gram,
                                   OrthogonalMatchingPursuit,
-                                  OrthogonalMatchingPursuitCV)
+                                  OrthogonalMatchingPursuitCV,
+                                  LinearRegression)
 from sklearn.utils.fixes import count_nonzero
+from sklearn.utils import check_random_state
 from sklearn.datasets import make_sparse_coded_signal
 
 n_samples, n_features, n_nonzero_coefs, n_targets = 20, 30, 5, 3
@@ -93,7 +95,6 @@ def test_bad_input():
 
 
 def test_perfect_signal_recovery():
-    # XXX: use signal generator
     idx, = gamma[:, 0].nonzero()
     gamma_rec = orthogonal_mp(X, y[:, 0], 5)
     gamma_gram = orthogonal_mp_gram(G, Xy[:, 0], 5)
@@ -218,3 +219,17 @@ def test_omp_cv():
                                     n_nonzero_coefs=ompcv.n_nonzero_coefs_)
     omp.fit(X, y_)
     assert_array_almost_equal(ompcv.coef_, omp.coef_)
+
+
+def test_omp_reaches_least_squares():
+    # Use small simple data; it's a sanity check but OMP can stop early
+    rng = check_random_state(0)
+    n_samples, n_features = (10, 8)
+    n_targets = 3
+    X = rng.randn(n_samples, n_features)
+    Y = rng.randn(n_samples, n_targets)
+    omp = OrthogonalMatchingPursuit(n_nonzero_coefs=n_features)
+    lstsq = LinearRegression()
+    omp.fit(X, Y)
+    lstsq.fit(X, Y)
+    assert_array_almost_equal(omp.coef_, lstsq.coef_)