Fix random state in algorithms (#234)

Author: wdevazelhes

metric_learn/lmnn.py

@@ -210,7 +210,7 @@ def fit(self, X, y):
       init = self.init
     self.components_ = _initialize_components(output_dim, X, y, init,
                                               self.verbose,
-                                              self.random_state)
+                                              random_state=self.random_state)
     required_k = np.bincount(label_inds).min()
     if self.k > required_k:
       raise ValueError('not enough class labels for specified k'

metric_learn/lsml.py

@@ -56,9 +56,8 @@ def _fit(self, quadruplets, weights=None):
     else:
       prior = self.prior
     M, prior_inv = _initialize_metric_mahalanobis(quadruplets, prior,
-                                                  return_inverse=True,
-                                                  strict_pd=True,
-                                                  matrix_name='prior')
+       return_inverse=True, strict_pd=True, matrix_name='prior',
+       random_state=self.random_state)
 
     step_sizes = np.logspace(-10, 0, 10)
     # Keep track of the best step size and the loss at that step.

metric_learn/nca.py

@@ -174,7 +174,8 @@ def fit(self, X, y):
       init = 'auto'
     else:
       init = self.init
-    A = _initialize_components(n_components, X, labels, init, self.verbose)
+    A = _initialize_components(n_components, X, labels, init, self.verbose,
+                               self.random_state)
 
     # Run NCA
     mask = labels[:, np.newaxis] == labels[np.newaxis, :]

metric_learn/sdml.py

@@ -69,9 +69,8 @@ def _fit(self, pairs, y):
     else:
       prior = self.prior
     _, prior_inv = _initialize_metric_mahalanobis(pairs, prior,
-                                                  return_inverse=True,
-                                                  strict_pd=True,
-                                                  matrix_name='prior')
+       return_inverse=True, strict_pd=True, matrix_name='prior',
+       random_state=self.random_state)
     diff = pairs[:, 0] - pairs[:, 1]
     loss_matrix = (diff.T * y).dot(diff)
     emp_cov = prior_inv + self.balance_param * loss_matrix

test/test_mahalanobis_mixin.py

@@ -652,3 +652,25 @@ def test_singular_array_init_or_prior(estimator, build_dataset, w0):
       with pytest.raises(LinAlgError) as raised_err:
         model.fit(input_data, labels)
       assert str(raised_err.value) == msg
+
+
+@pytest.mark.integration
+@pytest.mark.parametrize('estimator, build_dataset', metric_learners,
+                         ids=ids_metric_learners)
+def test_deterministic_initialization(estimator, build_dataset):
+  """Test that estimators that have a prior or an init are deterministic
+  when it is set to to random and when the random_state is fixed."""
+  input_data, labels, _, X = build_dataset()
+  model = clone(estimator)
+  if hasattr(estimator, 'init'):
+    model.set_params(init='random')
+  if hasattr(estimator, 'prior'):
+    model.set_params(prior='random')
+  model1 = clone(model)
+  set_random_state(model1, 42)
+  model1 = model1.fit(input_data, labels)
+  model2 = clone(model)
+  set_random_state(model2, 42)
+  model2 = model2.fit(input_data, labels)
+  np.testing.assert_allclose(model1.get_mahalanobis_matrix(),
+                             model2.get_mahalanobis_matrix())