| 198 | |
| 199 | |
| 200 | class BiasedWalker: |
| 201 | def __init__(self, idx2node, temp_path): |
| 202 | |
| 203 | self.idx2node = idx2node |
| 204 | self.idx = list(range(len(self.idx2node))) |
| 205 | self.temp_path = temp_path |
| 206 | pass |
| 207 | |
| 208 | def simulate_walks(self, num_walks, walk_length, stay_prob=0.3, workers=1, verbose=0): |
| 209 | |
| 210 | layers_adj = pd.read_pickle(self.temp_path + 'layers_adj.pkl') |
| 211 | layers_alias = pd.read_pickle(self.temp_path + 'layers_alias.pkl') |
| 212 | layers_accept = pd.read_pickle(self.temp_path + 'layers_accept.pkl') |
| 213 | gamma = pd.read_pickle(self.temp_path + 'gamma.pkl') |
| 214 | |
| 215 | nodes = self.idx # list(self.g.nodes()) |
| 216 | |
| 217 | results = Parallel(n_jobs=workers, verbose=verbose, )( |
| 218 | delayed(self._simulate_walks)(nodes, num, walk_length, stay_prob, layers_adj, layers_accept, layers_alias, |
| 219 | gamma) for num in |
| 220 | partition_num(num_walks, workers)) |
| 221 | |
| 222 | walks = list(itertools.chain(*results)) |
| 223 | return walks |
| 224 | |
| 225 | def _simulate_walks(self, nodes, num_walks, walk_length, stay_prob, layers_adj, layers_accept, layers_alias, gamma): |
| 226 | walks = [] |
| 227 | for _ in range(num_walks): |
| 228 | random.shuffle(nodes) |
| 229 | for v in nodes: |
| 230 | walks.append(self._exec_random_walk(layers_adj, layers_accept, layers_alias, |
| 231 | v, walk_length, gamma, stay_prob)) |
| 232 | return walks |
| 233 | |
| 234 | def _exec_random_walk(self, graphs, layers_accept, layers_alias, v, walk_length, gamma, stay_prob=0.3): |
| 235 | initialLayer = 0 |
| 236 | layer = initialLayer |
| 237 | |
| 238 | path = [] |
| 239 | path.append(self.idx2node[v]) |
| 240 | |
| 241 | while len(path) < walk_length: |
| 242 | r = random.random() |
| 243 | if (r < stay_prob): # same layer |
| 244 | v = chooseNeighbor(v, graphs, layers_alias, |
| 245 | layers_accept, layer) |
| 246 | path.append(self.idx2node[v]) |
| 247 | else: # different layer |
| 248 | r = random.random() |
| 249 | try: |
| 250 | x = math.log(gamma[layer][v] + math.e) |
| 251 | p_moveup = (x / (x + 1)) |
| 252 | except: |
| 253 | print(layer, v) |
| 254 | raise ValueError() |
| 255 | |
| 256 | if (r > p_moveup): |
| 257 | if (layer > initialLayer): |