FIX PCA error handling for invalid n_components

larsmans · larsmans · commit f309126b534f · 2014-06-05T20:16:43.000+02:00
n_components &lt; 0 was silently handled like None; &gt; n_features gave a
cryptic error message from the bowels of np.linalg.

XXX n_components == 0 is still allowed as changing it makes tests fail,
not sure what it means.
diff --git a/sklearn/decomposition/pca.py b/sklearn/decomposition/pca.py
@@ -292,6 +292,9 @@ def _fit(self, X):
 
             n_components = _infer_dimension_(explained_variance_,
                                              n_samples, n_features)
+        elif not 0 <= n_components <= n_features:
+            raise ValueError("n_components=%r invalid for n_features=%d"
+                             % (n_components, n_features))
 
         if 0 < n_components < 1.0:
             # number of components for which the cumulated explained variance
diff --git a/sklearn/decomposition/tests/test_pca.py b/sklearn/decomposition/tests/test_pca.py
@@ -6,6 +6,7 @@
 from sklearn.utils.testing import assert_true
 from sklearn.utils.testing import assert_equal
 from sklearn.utils.testing import assert_greater
+from sklearn.utils.testing import assert_raises
 from sklearn.utils.testing import assert_warns
 
 from sklearn import datasets
@@ -165,6 +166,12 @@ def test_pca_inverse():
     assert_almost_equal(relative_max_delta, 0.11, decimal=2)
 
 
+def test_pca_validation():
+    X = [[0, 1], [1, 0]]
+    for n_components in [-1, 3]:
+        assert_raises(ValueError, PCA(n_components).fit, X)
+
+
 def test_randomized_pca_check_projection():
     """Test that the projection by RandomizedPCA on dense data is correct"""
     rng = np.random.RandomState(0)
