swev-id: scikit-learn__scikit-learn-13142 align GaussianMixture fit_predict with predict

sklearn/mixture/base.py

@@ -257,11 +257,6 @@ def fit_predict(self, X, y=None):
                 best_params = self._get_parameters()
                 best_n_iter = n_iter
 
-        # Always do a final e-step to guarantee that the labels returned by
-        # fit_predict(X) are always consistent with fit(X).predict(X)
-        # for any value of max_iter and tol (and any random_state).
-        _, log_resp = self._e_step(X)
-
         if not self.converged_:
             warnings.warn('Initialization %d did not converge. '
                           'Try different init parameters, '
@@ -273,7 +268,9 @@ def fit_predict(self, X, y=None):
         self.n_iter_ = best_n_iter
         self.lower_bound_ = max_lower_bound
 
-        return log_resp.argmax(axis=1)
+        # Compute labels through the same path as predict to keep outputs
+        # identical between fit_predict(X) and fit(X).predict(X).
+        return self.predict(X)
 
     def _e_step(self, X):
         """E step.

sklearn/mixture/tests/test_gaussian_mixture.py

@@ -198,6 +198,49 @@ def test_check_X():
     assert_array_equal(X, _check_X(X, n_components, n_features))
 
 
+@pytest.mark.parametrize("seed", [0, 1, 2])
+def test_fit_predict_matches_predict_across_seeds(seed):
+    rng = np.random.RandomState(seed)
+    X = rng.randn(200, 3)
+
+    gmm = GaussianMixture(
+        n_components=3,
+        n_init=5,
+        covariance_type='full',
+        random_state=seed,
+    )
+
+    labels_fit_predict = gmm.fit_predict(X)
+    labels_predict = gmm.predict(X)
+
+    assert_array_equal(labels_fit_predict, labels_predict)
+
+
+def test_fit_predict_uses_best_init():
+    rng = np.random.RandomState(0)
+    X = rng.randn(120, 2)
+
+    gmm_fit_predict = GaussianMixture(
+        n_components=2,
+        n_init=5,
+        covariance_type='full',
+        random_state=0,
+    )
+
+    labels_fit_predict = gmm_fit_predict.fit_predict(X)
+
+    gmm_fit_then_predict = GaussianMixture(
+        n_components=2,
+        n_init=5,
+        covariance_type='full',
+        random_state=0,
+    )
+
+    labels_predict = gmm_fit_then_predict.fit(X).predict(X)
+
+    assert_array_equal(labels_fit_predict, labels_predict)
+
+
 def test_check_weights():
     rng = np.random.RandomState(0)
     rand_data = RandomData(rng)