add xnetmf

Author: kinkunchan

.DS_Store

@@ -204,20 +204,20 @@ def get_representations(graph, rep_method, verbose = True):
 		reprsn = reprsn / np.linalg.norm(reprsn, axis = 1).reshape((reprsn.shape[0],1))
 	return reprsn
 
-if __name__ == "__main__":
-	if len(sys.argv) < 2:
-		#####PUT IN YOUR GRAPH AS AN EDGELIST HERE (or pass as cmd line argument)#####  
-		#(see networkx read_edgelist() method...if networkx can read your file as an edgelist you're good!)
-		graph_file = "data/arenas_combined_edges.txt"
-	else:
-		graph_file = sys.argv[1]
-	nx_graph = nx.read_edgelist(graph_file, nodetype = int, comments="%")
-	adj_matrix = nx.adjacency_matrix(nx_graph).todense()
+# if __name__ == "__main__":
+# 	if len(sys.argv) < 2:
+# 		#####PUT IN YOUR GRAPH AS AN EDGELIST HERE (or pass as cmd line argument)#####  
+# 		#(see networkx read_edgelist() method...if networkx can read your file as an edgelist you're good!)
+# 		graph_file = "data/arenas_combined_edges.txt"
+# 	else:
+# 		graph_file = sys.argv[1]
+# 	nx_graph = nx.read_edgelist(graph_file, nodetype = int, comments="%")
+# 	adj_matrix = nx.adjacency_matrix(nx_graph).todense()
 
-	graph = Graph(adj_matrix)
-	rep_method = RepMethod(max_layer = 2) #Learn representations with xNetMF.  Can adjust parameters (e.g. as in REGAL)
-	representations = get_representations(graph, rep_method)
-	print(representations.shape)
+# 	graph = Graph(adj_matrix)
+# 	rep_method = RepMethod(max_layer = 2) #Learn representations with xNetMF.  Can adjust parameters (e.g. as in REGAL)
+# 	representations = get_representations(graph, rep_method)
+# 	print(representations.shape)
 
 
 

semb/.DS_Store

@@ -13,10 +13,11 @@ def train(self):
         # learn representations with xNetMF.  Can adjust parameters (e.g. as in REGAL)
         rep_method = RepMethod(max_layer=self.params['max_layer'], p=self.params['dim'], 
                               alpha=self.params['discount'], gammastruc=self.params['gamma']) 
-        # FIXME: this doesnt look like a standard embeddings format
-        representations = get_representations(self.graph, rep_method)
-        list_nodes = self.graph.node()
+        graph = Graph(nx.adjacency_matrix(self.graph))
+        representations = get_representations(graph, rep_method)
+
         self.embeddings = dict()
-        for i in range(0, len(list_nodes)):
-        	self.embeddings[i] = representations[i].tolist()
+        list_nodes = list(self.graph.nodes())
 
+        for i in range(0, representations.shape[0]):
+            self.embeddings[list_nodes[i]] = representations[i].tolist()

semb/methods/xnetmf/internal/xnetmf.py

@@ -106,7 +106,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 3,
+   "execution_count": 4,
    "metadata": {
     "tags": []
    },
@@ -115,20 +115,37 @@
      "name": "stdout",
      "output_type": "stream",
      "text": [
-      "rm /Users/mark/GoogleDrive/UM/S4/GEMS/Git/StrucEmbeddingLibrary/semb/methods/struc2vec/pickles/weights_distances-layer-*.pickle\n"
+      "max degree:  80\n",
+      "got k hop neighbors in time:  0.0337069034576416\n",
+      "got degree sequences in time:  0.013570070266723633\n",
+      "computed representation in time:  0.010637044906616211\n",
+      "[[-1.37432577e-04  1.30331318e-03 -2.69511558e-04 ... -2.44950932e-22\n",
+      "   1.63093975e-22  1.42838586e-22]\n",
+      " [-9.78507622e-05  1.49474244e-04 -2.42289473e-04 ... -2.88146479e-22\n",
+      "   2.12115482e-22  1.59210219e-22]\n",
+      " [-1.10454466e-05 -3.63693553e-05  1.05660366e-04 ... -2.95114433e-22\n",
+      "   2.13269839e-22  1.64978283e-22]\n",
+      " ...\n",
+      " [-5.87339184e-05 -1.25072095e-04 -8.59305162e-06 ... -5.85895127e-22\n",
+      "   4.11111734e-22  1.68655414e-22]\n",
+      " [-5.15959810e-05 -1.13438423e-04 -1.00442930e-06 ... -5.89756433e-22\n",
+      "   4.14443476e-22  1.70109439e-22]\n",
+      " [ 2.61171369e-04 -9.51229394e-05 -4.38703782e-05 ... -5.83886813e-22\n",
+      "   4.11219006e-22  1.66800475e-22]]\n",
+      "<semb.methods.xnetmf.internal.config.Graph object at 0x7fa67c892fd0>\n"
      ]
     }
    ],
    "source": [
     "from semb.methods import load as load_method\n",
     "from semb.datasets import load as load_dataset\n",
     "\n",
-    "Node2VecMethod = load_method(\"struc2vec\")\n",
+    "Node2VecMethod = load_method(\"xnetmf\")\n",
     "AirportDataProvider = load_dataset(\"airports\")\n",
     "airport_datasets = AirportDataProvider().get_datasets()\n",
     "brazil_airport_graph = AirportDataProvider().load_dataset(airport_datasets[0])\n",
     "\n",
-    "node2vec = Node2VecMethod(brazil_airport_graph, opt1=True, opt2=True, opt3=True)\n",
+    "node2vec = Node2VecMethod(brazil_airport_graph)\n",
     "node2vec.train()\n",
     "emb = node2vec.get_embeddings()"
    ]