Disabling non deterministic tests in test_n2v

See issue #880

tests/embed/test_n2v.py

@@ -12,68 +12,70 @@
 
 
 class Node2VecEmbedTest(unittest.TestCase):
-    def test_n2v_returns_same_labels_with_different_nodeid_types(self):
-        probability_matrix = np.array([[0.95, 0.01], [0.01, 0.95]])
-        number_of_nodes_per_community = [20, 20]
+    # these are not deterministic tests and can't be trusted to work consistently.
+    # See Issue #880
+#     def test_n2v_returns_same_labels_with_different_nodeid_types(self):
+#         probability_matrix = np.array([[0.95, 0.01], [0.01, 0.95]])
+#         number_of_nodes_per_community = [20, 20]
 
-        sbm_sample = gc.simulations.sbm(
-            number_of_nodes_per_community, probability_matrix
-        )
-        sbm_graph = nx.from_numpy_array(sbm_sample)
+#         sbm_sample = gc.simulations.sbm(
+#             number_of_nodes_per_community, probability_matrix
+#         )
+#         sbm_graph = nx.from_numpy_array(sbm_sample)
 
-        graph = nx.Graph()
-        graph_as_strings = nx.Graph()
-        for s, t in sbm_graph.edges():
-            graph.add_edge(s, t, weight=1)
-            graph_as_strings.add_edge(str(s), str(t), weight=1)
+#         graph = nx.Graph()
+#         graph_as_strings = nx.Graph()
+#         for s, t in sbm_graph.edges():
+#             graph.add_edge(s, t, weight=1)
+#             graph_as_strings.add_edge(str(s), str(t), weight=1)
 
-        original_embedding = gc.embed.node2vec_embed(graph, random_seed=1)
-        string_embedding = gc.embed.node2vec_embed(graph_as_strings, random_seed=1)
+#         original_embedding = gc.embed.node2vec_embed(graph, random_seed=1)
+#         string_embedding = gc.embed.node2vec_embed(graph_as_strings, random_seed=1)
 
-        k = KMeans(n_clusters=2)
-        original_labels = k.fit_predict(original_embedding[0])
-        string_labels = k.fit_predict(string_embedding[0])
+#         k = KMeans(n_clusters=2)
+#         original_labels = k.fit_predict(original_embedding[0])
+#         string_labels = k.fit_predict(string_embedding[0])
 
-        expected_labels = np.zeros(40, dtype=int)
-        expected_labels[20:] = 1
+#         expected_labels = np.zeros(40, dtype=int)
+#         expected_labels[20:] = 1
 
-        original_ari = adjusted_rand_score(original_labels, expected_labels)
-        string_ari = adjusted_rand_score(string_labels, expected_labels)
+#         original_ari = adjusted_rand_score(original_labels, expected_labels)
+#         string_ari = adjusted_rand_score(string_labels, expected_labels)
 
-        self.assertEqual(original_ari, string_ari)
+#         self.assertEqual(original_ari, string_ari)
 
-    def test_n2v_directed_undirected_returns_same_clustering(self):
-        probability_matrix = np.array([[0.95, 0.01], [0.01, 0.95]])
-        number_of_nodes_per_community = [20, 20]
+#     def test_n2v_directed_undirected_returns_same_clustering(self):
+#         probability_matrix = np.array([[0.95, 0.01], [0.01, 0.95]])
+#         number_of_nodes_per_community = [20, 20]
 
-        sbm_sample = gc.simulations.sbm(
-            number_of_nodes_per_community, probability_matrix
-        )
-        sbm_graph = nx.from_numpy_array(sbm_sample)
+#         sbm_sample = gc.simulations.sbm(
+#             number_of_nodes_per_community, probability_matrix
+#         )
+#         sbm_graph = nx.from_numpy_array(sbm_sample)
 
-        graph = nx.Graph()
-        graph_directed = nx.DiGraph()
-        for s, t in sbm_graph.edges():
-            graph.add_edge(s, t, weight=1)
+#         graph = nx.Graph()
+#         graph_directed = nx.DiGraph()
+#         for s, t in sbm_graph.edges():
+#             graph.add_edge(s, t, weight=1)
 
-            graph_directed.add_edge(s, t, weight=1)
-            graph_directed.add_edge(t, s, weight=1)
+#             graph_directed.add_edge(s, t, weight=1)
+#             graph_directed.add_edge(t, s, weight=1)
 
-        undirected_embedding = gc.embed.node2vec_embed(graph, random_seed=1)
-        directed_embedding = gc.embed.node2vec_embed(graph_directed, random_seed=1)
+#         undirected_embedding = gc.embed.node2vec_embed(graph, random_seed=1)
+#         directed_embedding = gc.embed.node2vec_embed(graph_directed, random_seed=1)
 
-        k = KMeans(n_clusters=2, random_state=1234)
-        undirected_labels = k.fit_predict(undirected_embedding[0])
-        k = KMeans(n_clusters=2, random_state=1234)
-        directed_labels = k.fit_predict(directed_embedding[0])
+#         k = KMeans(n_clusters=2, random_state=1234)
+#         undirected_labels = k.fit_predict(undirected_embedding[0])
+#         k = KMeans(n_clusters=2, random_state=1234)
+#         directed_labels = k.fit_predict(directed_embedding[0])
 
-        expected_labels = np.zeros(40, dtype=int)
-        expected_labels[20:] = 1
+#         expected_labels = np.zeros(40, dtype=int)
+#         expected_labels[20:] = 1
 
-        undirected_ari = adjusted_rand_score(undirected_labels, expected_labels)
-        directed_ari = adjusted_rand_score(directed_labels, expected_labels)
+#         undirected_ari = adjusted_rand_score(undirected_labels, expected_labels)
+#         directed_ari = adjusted_rand_score(directed_labels, expected_labels)
 
-        self.assertEqual(undirected_ari, directed_ari)
+#         self.assertEqual(undirected_ari, directed_ari)
 
     def test_node2vec_embed(self):
         g = nx.florentine_families_graph()