scikit-learn
diff --git a/‎sklearn/decomposition/nmf.py
+16-4 b/‎sklearn/decomposition/nmf.py
+16-4
diff --git a/‎sklearn/decomposition/tests/test_nmf.py
+32 b/‎sklearn/decomposition/tests/test_nmf.py
+32
@@ -12,6 +12,7 @@
 
 from math import sqrt
 import warnings
+import numbers
 
 import numpy as np
 import scipy.sparse as sp
@@ -60,10 +61,11 @@ def check_non_negative(X, whom):
 def _check_init(A, shape, whom):
     A = check_array(A)
     if np.shape(A) != shape:
-        raise ValueError('TODO')
+        raise ValueError('Array with wrong shape passed to %s. Expected %s, '
+                         'but got %s ' % (whom, shape, np.shape(A)))
     check_non_negative(A, whom)
     if np.max(A) == 0:
-        raise ValueError('TODO')
+        raise ValueError('Array passed to %s is full of zeros.' % whom)
 
 
 def _safe_compute_error(X, W, H):
@@ -227,7 +229,8 @@ def _initialize_nmf(X, n_components, init=None, eps=1e-6,
     else:
         raise ValueError(
             'Invalid init parameter: got %r instead of one of %r' %
-            (init, (None, 'nndsvd', 'nndsvda', 'nndsvdar', 'random')))
+            (init, (None, 'random', 'nndsvd', 'nndsvda', 'nndsvdar',
+                    'uniform')))
 
     return W, H
 
@@ -680,6 +683,16 @@ def non_negative_matrix_factorization(X, W=None, H=None, n_components=None,
     if n_components is None:
         n_components = n_features
 
+    if not isinstance(n_components, int) or n_components <= 0:
+        raise ValueError("Number of components must be positive;"
+                         " got (n_components=%r)" % n_components)
+    if not isinstance(max_iter, numbers.Number) or max_iter < 0:
+        raise ValueError("Maximum number of iteration must be positive;"
+                         " got (max_iter=%r)" % max_iter)
+    if not isinstance(tol, numbers.Number) or tol < 0:
+        raise ValueError("Tolerance for stopping criteria must be "
+                         "positive; got (tol=%r)" % tol)
+
     # check W and H, or initialize them
     if init == 'custom':
         _check_init(H, (n_components, n_features), "NMF (input H)")
@@ -920,7 +933,6 @@ def transform(self, X):
             Transformed data
         """
         check_is_fitted(self, 'n_components_')
-        X = check_array(X, accept_sparse='csc')
 
         W, _, _ = non_negative_matrix_factorization(
             X=X, W=None, H=self.components_, n_components=self.n_components_,
 
@@ -177,3 +177,35 @@ def test_sparse_transform():
         A_fit_tr = model.fit_transform(A)
         A_tr = model.transform(A)
         assert_array_almost_equal(A_fit_tr, A_tr, decimal=2)
+
+
+def test_non_negative_matrix_factorization_path():
+    # Test path consistency between the class and the public function
+    A = np.abs(random_state.randn(10, 10))
+    A[:, 2 * np.arange(5)] = 0
+
+    for solver in ('proj-grad', 'coordinate'):
+        W_nmf, H, _ = nmf.non_negative_matrix_factorization(
+            A, solver=solver, random_state=1, tol=1e-2)
+        W_nmf_bis, H, _ = nmf.non_negative_matrix_factorization(
+            A, H=H, update_H=False, solver=solver, random_state=1, tol=1e-2)
+
+        model_class = nmf.NMF(solver=solver, random_state=1, tol=1e-2)
+        W_cls = model_class.fit_transform(A)
+        W_cls_bis = model_class.transform(A)
+        assert_array_almost_equal(W_nmf, W_cls, decimal=10)
+        assert_array_almost_equal(W_nmf_bis, W_cls_bis, decimal=10)
+
+
+def test_non_negative_matrix_factorization_checking():
+    A = np.ones((2, 2))
+    # Test parameters checking is public function
+    nnmf = nmf.non_negative_matrix_factorization
+    msg = "Number of components must be positive; got (n_components='2')"
+    assert_raise_message(ValueError, msg, nnmf, A, A, A, '2')
+    msg = "Negative values in data passed to NMF (input H)"
+    assert_raise_message(ValueError, msg, nnmf, A, A, -A, 2, 'custom')
+    msg = "Negative values in data passed to NMF (input W)"
+    assert_raise_message(ValueError, msg, nnmf, A, -A, A, 2, 'custom')
+    msg = "Array passed to NMF (input H) is full of zeros"
+    assert_raise_message(ValueError, msg, nnmf, A, A, 0 * A, 2, 'custom')