fix struc2vec

Author: kinkunchan

.DS_Store

@@ -147,3 +147,6 @@ cython_debug/
 # Mac OS file
 *.DS_Store
 
+# struc2vec pickles
+semb/methods/struc2vec/pickles/
+

.gitignore

@@ -29,7 +29,7 @@ def train(self):
         model = Word2Vec(walks, size=self.params['dim'], window=self.params['window_size'],
                          min_count=0, sg=1, workers=self.params['workers'], iter=self.params['iter'])
         self.model = model
-        print(g.G)
+        # print(g.G)
         self.embeddings = dict()
         for cur_node in g.G.nodes():
             self.embeddings[cur_node] = self.model.wv.get_vector(str(cur_node)).tolist()

semb/.DS_Store

@@ -185,7 +185,7 @@ def generate_random_walks(num_walks,walk_length,workers,vertices):
     save_random_walks(walks)
 
 def save_random_walks(walks):
-    with open('random_walks.txt', 'w') as file:
+    with open(walk_fname, 'w') as file:
         for walk in walks:
             line = ''
             for v in walk:

semb/methods/node2vec/method.py

@@ -607,7 +607,7 @@ def generate_distances_network(workers):
     logging.info('Creating distance network...')
 
     os.system("rm "+returnPickle().replace(' ', '\ ')+"weights_distances-layer-*.pickle")
-    print('!!!!!')
+    # print('!!!!!')
     print("rm "+returnPickle().replace(' ', '\ ')+"weights_distances-layer-*.pickle")
     with ProcessPoolExecutor(max_workers=1) as executor:
         job = executor.submit(generate_distances_network_part1,workers)

semb/methods/struc2vec/algorithms.py

@@ -229,7 +229,7 @@ def load_matfile(file_, variable_name="network", undirected=True):
 def from_networkx(G_input, undirected=True):
     G = Graph()
 
-    for idx, x in enumerate(G_input.nodes_iter()):
+    for idx, x in enumerate(G_input.nodes()):
         for y in iterkeys(G_input[x]):
             G[x].append(y)
 

semb/methods/struc2vec/algorithms_distances.py

@@ -5,48 +5,56 @@
 from gensim.models.word2vec import LineSentence
 from time import time
 from . import graph
+from .utils import *
+from .graph import *
 from semb.methods import BaseMethod
 
 """
 Driver for learning Struc2Vec embedding
 """
 class Method(BaseMethod):
 
-	__PARAMS__ = dict(dim=128, walk_length=80, num_walks=10, window_size=10, util_layer=None, \
-					  iter=5, worker=1, weighted=False, directed=False, opt1=False, opt2=False, opt3=False)
-
-
-	def exec_struc2vec(self):
-		# execute struc2vec
-		if self.params['opt3']:
-			until_layer = self.params['until_layer']
-		else:
-			until_layer = None
-		G = struc2vec.Graph(self.graph, self.params['directed'],
-		                    self.params['worker'], untilLayer=until_layer)
-		if self.params['opt1']:
-			G.preprocess_neighbors_with_bfs_compact()
-		else:
-			G.preprocess_neighbors_with_bfs()
-		if self.params['opt2']:
-			G.create_vectors()
-			G.calc_distances(compactDegree=self.params['opt1'])
-		else:
-			G.calc_distances_all_vertices(compactDegree=self.params['opt1'])
-
-		G.create_distances_network()
-		G.preprocess_parameters_random_walk()
-		G.simulate_walks(self.params['num_walks'], self.params['walk_length'])
-
-	def learn_embeddings(self):
-		walks = LineSentence('random_walks.txt')
-		self.model = Word2Vec(walks, size=self.params['dim'], window=self.params['window_size'],
-	                	 min_count=0, hs=1, sg=1, workers=self.params['worker'], iter=self.params['iter'])
-		self.embeddings = self.model.wv
-
-	def get_id(self):
-		return "struc2vec"
-
-	def train(self):
-		self.exec_struc2vec()
-		self.learn_embeddings()
+    __PARAMS__ = dict(dim=128, walk_length=80, num_walks=10, window_size=10, until_layer=None, \
+                      iter=5, workers=1, weighted=False, directed=False, opt1=False, opt2=False, opt3=False)
+
+
+    def exec_struc2vec(self):
+        # execute struc2vec
+        if self.params['opt3']:
+            until_layer = self.params['until_layer']
+        else:
+            until_layer = None
+        graph = from_networkx(self.graph, not self.params['directed'])
+        G = struc2vec.Graph(graph, self.params['directed'], self.params['workers'], untilLayer=until_layer)
+        
+        if self.params['opt1']:
+            G.preprocess_neighbors_with_bfs_compact()
+        else:
+            G.preprocess_neighbors_with_bfs()
+        if self.params['opt2']:
+            G.create_vectors()
+            G.calc_distances(compactDegree=self.params['opt1'])
+        else:
+            G.calc_distances_all_vertices(compactDegree=self.params['opt1'])
+
+        G.create_distances_network()
+        G.preprocess_parameters_random_walk()
+        G.simulate_walks(self.params['num_walks'], self.params['walk_length'])
+
+
+
+    def learn_embeddings(self):
+        walks = LineSentence(walk_fname)
+        self.model = Word2Vec(walks, size=self.params['dim'], window=self.params['window_size'],
+                         min_count=0, hs=1, sg=1, workers=self.params['workers'], iter=self.params['iter'])
+
+        self.embeddings = dict()
+        for cur_node in self.graph.nodes():
+            self.embeddings[cur_node] = self.model.wv.get_vector(str(cur_node)).tolist()
+
+    def get_id(self):
+        return "struc2vec"
+
+    def train(self):
+        self.exec_struc2vec()
+        self.learn_embeddings()