-
Notifications
You must be signed in to change notification settings - Fork 2
/
GraphM.py
449 lines (375 loc) · 18.8 KB
/
GraphM.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
import itertools
import math
from tqdm import tqdm
import torch
import time
import scipy.sparse as sp
from torch_geometric.datasets import Planetoid
from scipy.sparse import csr_matrix, save_npz, load_npz
import networkx as nx
import os
import numpy as np
from dataset.homo_data.planetoid import Planetoid
from dataset.homo_data.ogbn import Ogbn
from dataset.homo_data.ogbn_100m import Ogbn_papers100m
from dataset.homo_data.flickr import Flickr
from dataset.homo_data.reddit import Reddit
from dataset.homo_data.ppi_small import PPI_small
from dataset.homo_data.amazon import Amazon
from dataset.homo_data.coauthor import Coauthor
from utils import label_node_homogeneity
import matplotlib.pyplot as plt
class GraphData:
def __init__(self, root, name, args):
self.data_name = name
self.args = args
self.root = os.path.join(root)
self.load_homo_simplex_dataset(name=args.data_name, root=args.root,
split=args.split)
# self.load_data()
self.dataset.mining_matrix = None
self.dataset.mining_list = None
def node_homo(self):
G = nx.Graph()
G.add_nodes_from(range(self.dataset.data.num_node))
G.add_edges_from(zip(self.dataset.data.edge.col.tolist(),
self.dataset.data.edge.row.tolist()))
degrees = dict(G.degree())
max_degree = max(degrees.values())
boundaries = []
for i in range(10, 101, 10):
# print(i)
boundaries.append(int(max_degree*i/100))
result = []
prev_boundary = 0
for boundary in boundaries:
subset = [i for i, value in enumerate(
degrees) if prev_boundary <= value <= boundary]
result.append(subset)
prev_boundary = boundary
node_homogeneity = []
for l in result:
homogeneity = label_node_homogeneity(
G=self.dataset.data, node_index=l)
node_homogeneity.append(homogeneity)
# 绘制图表
x_ticks = [f"{i}-{i + 10}%" for i in range(0, 101, 10)]
x = np.arange(len(x_ticks))
y1 = [len(l) for l in result]
print(y1)
y2 = node_homogeneity
fig, ax1 = plt.subplots()
ax2 = ax1.twinx()
ax1.bar(x, y1, alpha=0.7, color='b', align='center')
ax2.plot(x, y2, '-o', color='r')
ax1.set_xlabel('Percentage of Nodes')
ax1.set_ylabel('Percentage of Nodes in Subset', color='b')
ax2.set_ylabel('Average Node Homophily', color='r')
ax1.set_xticks(x)
ax1.set_xticklabels(x_ticks)
plt.title('Node Subset Analysis')
plt.show()
exit(0)
# homo = label_node_homogeneity(G=self.dataset.data)
# print("the average homo is {}".format(homo))
def calculate_node_homo(self):
num_nodes = self.dataset.data.num_node
homophily_node = 0
for edge_u in tqdm(range(num_nodes)):
hit = 0
# 遍历所有节点，通过边找出他们的邻居
edge_v_list = self.dataset.data.edge.col[torch.where(
torch.tensor(self.dataset.data.edge.row) == edge_u)].tolist()
if isinstance(edge_v_list, list) and len(edge_v_list) != 0:
for i in range(len(edge_v_list)):
edge_v = edge_v_list[i]
if self.dataset.data.y[edge_u] == self.dataset.data.y[edge_v]:
hit += 1
homophily_node += hit / len(edge_v_list)
else:
if self.dataset.data.y[edge_u] == self.dataset.data.y[edge_v_list]:
hit += 1
homophily_node += hit
homophily_node /= num_nodes
print(
"the node-level homophily of {} is: {}".format(self.data_name, homophily_node))
num_edges = len(self.dataset.data.edge.row)
homophily_edge = 0
for i in tqdm(range(num_edges)):
if self.dataset.data.y[self.dataset.data.edge.row[i]] == self.dataset.data.y[self.dataset.data.edge.col[i]]:
homophily_edge += 1
homophily_edge /= num_edges
print(
"the edge-level homophily of {} is: {}".format(self.data_name, homophily_edge))
with open('output.txt', 'a') as f:
f.write(f"{self.dataset.name}\n ")
f.write(
"the node-level homophily of {} is: {}".format(self.data_name, homophily_node))
f.write("\n")
f.write(
"the edge-level homophily of {} is: {}".format(self.data_name, homophily_edge))
f.write("\n")
exit(0)
def drop_nodes(self, ratio=0.1):
try:
loaded_matrix = load_npz(os.path.join("/mnt/ssd2/home/xkli/mjy",
'{} by {} adj_matrix.npz'.format(self.data_name, ratio)))
self.dataset.adj = loaded_matrix
print("already has the drop node matrix!")
except:
print("drop node base on degree, ratio {}".format(ratio))
G = nx.Graph()
G.add_nodes_from(range(self.dataset.data.num_node))
G.add_edges_from(zip(self.dataset.data.edge.col.tolist(),
self.dataset.data.edge.row.tolist()))
degrees = dict(G.degree())
print(degrees)
# drop训练集
# sorted_nodes = sorted(degrees, key=degrees.get, reverse=True)
# top_10_percent = int(len(self.dataset.train_idx) * ratio)
# top_nodes = []
# i = 0
# for u in sorted_nodes:
# if u in self.dataset.train_idx:
# top_nodes.append(u)
# i += 1
# if i == top_10_percent:
# break
# drop所有
sorted_nodes = sorted(degrees, key=degrees.get, reverse=True)
top_10_percent = int(self.dataset.data.num_node * ratio)
top_nodes = []
i = 0
for u in sorted_nodes:
top_nodes.append(u)
i += 1
if i == top_10_percent:
break
# top_nodes = sorted_nodes[:top_10_percent]
# print(top_nodes)
# for u, v in tqdm(zip(self.dataset.data.edge.row.tolist(), self.dataset.data.edge.col.tolist())):
# # print(u, v)
# if (u in top_nodes) or (v in top_nodes):
# pass
# else:
# row_new.append(u)
# col_new.append(v)
# weight_new.append(1)
# self.dataset.adj = csr_matrix((weight_new, (row_new, col_new)), shape=(
# self.dataset.data.num_node, self.dataset.data.num_node))
# save_npz(os.path.join("/mnt/ssd2/home/xkli/mjy",
# '{} by {} adj_matrix.npz'.format(self.data_name, ratio)), self.dataset.adj)
for u in tqdm(top_nodes):
for v in set(self.dataset.adj.getrow(u).indices):
row_i = self.dataset.adj.indptr[u]
row_i_next = self.dataset.adj.indptr[u + 1]
row_j = self.dataset.adj.indptr[v]
row_j_next = self.dataset.adj.indptr[v + 1]
pos_ij = np.searchsorted(
self.dataset.adj.indices[row_i:row_i_next], v) + row_i
pos_ji = np.searchsorted(
self.dataset.adj.indices[row_j:row_j_next], u) + row_j
# 本来有边直接改
if pos_ij < row_i_next and self.dataset.adj.indices[pos_ij] == v:
self.dataset.adj.data[pos_ij] = 0
self.dataset.adj.data[pos_ji] = 0
t = time.time()
indices_with_zero_weight = np.where(self.dataset.adj.data == 0)[0]
self.dataset.adj.data = np.delete(
self.dataset.adj.data, indices_with_zero_weight)
self.dataset.adj.indices = np.delete(
self.dataset.adj.indices, indices_with_zero_weight)
self.dataset.adj.indptr[1:] -= np.searchsorted(
indices_with_zero_weight, self.dataset.adj.indptr[1:])
save_npz(os.path.join("/mnt/ssd2/home/xkli/mjy",
'{} by {} adj_matrix.npz'.format(self.data_name, ratio)), self.dataset.adj)
print(time.time()-t)
exit(0)
# print(self.dataset.adj)
# self.dataset.data.adj = sp.coo_matrix((torch.ones([len(self.dataset.data.edge.row)]),
# (row_new, col_new)),
# shape=(self.dataset.data.num_node, self.dataset.data.num_node))
def load_homo_simplex_dataset(self, name, root, split):
if name.lower() in ('cora', 'citeseer', 'pubmed'):
dataset = Planetoid(name, root, split)
self.root = os.path.join(root, "Planetoid")
if name.lower() in ('arxiv', 'products'):
dataset = Ogbn(name, root, split)
self.root = os.path.join(root, "ogbn")
if name.lower() in ('flickr'):
dataset = Flickr(name, root, split)
self.root = os.path.join(root, "Flickr")
if name.lower() in ('reddit'):
dataset = Reddit(name, root, split)
self.root = os.path.join(root, "Reddit")
if name.lower() in ('papers100m'):
dataset = Ogbn_papers100m()
self.root = os.path.join(root, "papers100M")
if name.lower() in ('ppi_small'):
dataset = PPI_small(name, root, split)
self.root = os.path.join(root, "ppi_small")
if name.lower() in ('computers', 'photo'):
dataset = Amazon(name, root, split)
self.root = os.path.join(root, "amazon")
if name.lower() in ('cs', 'phy'):
dataset = Coauthor(name, root, split)
self.root = os.path.join(root, "coauthor")
self.dataset = dataset
self.dataset.data.adj = sp.coo_matrix((torch.ones([len(self.dataset.data.edge.row)]),
(self.dataset.data.edge.row, self.dataset.data.edge.col)),
shape=(self.dataset.data.num_node, self.dataset.data.num_node))
return dataset
def mining(self, method, a=0.5, b=0.5, c=0.5):
print("Begin to mining the data!")
start_time = time.time()
if method is None:
raise NotImplemented
elif method in [ "Engienvector_centrality",
"clustering_coefficients",
"degree_centrality", "together", ]:
try:
if method == "together":
centrality1 = torch.load(os.path.join(
self.root, self.data_name, "degree_centrality"))
centrality2 = torch.load(os.path.join(
self.root, self.data_name, "clustering_coefficients"))
try:
centrality3 = torch.load(os.path.join(
self.root, self.data_name, "Engienvector_centrality"))
except:
print("there is no engienvector for the current dataset!")
centrality3 = torch.load(os.path.join(
self.root, self.data_name, "degree_centrality"))
c = 0
for i, centrality in enumerate([centrality1, centrality2, centrality3]):
num_outliers = int(0.05 * centrality.size()[0])
arr = centrality
_, indices = torch.topk(arr.abs(), num_outliers)
outliers = arr[indices]
arr[indices] = 1.0
min_value = arr.min()
max_value = arr[arr != 1.0].max()
centrality = (arr - min_value) / \
(max_value - min_value)
centrality[arr == 1.0] = 1.0
if i == 0:
centrality1 = centrality
if i == 1:
centrality2 = centrality
if i == 2:
centrality3 = centrality
centrality = a*centrality1 + b*centrality2 + c*centrality3
min_val = torch.min(centrality)
max_val = torch.max(centrality)
centrality = (centrality - min_val) / (max_val - min_val)
self.dataset.mining_list = centrality
print("already has the required matrix")
return centrality
else:
centrality = torch.load(os.path.join(
self.root, self.data_name, method))
min_val = torch.min(centrality)
max_val = torch.max(centrality)
centrality = (centrality - min_val) / (max_val - min_val)
self.dataset.mining_list = centrality
centrality1 = torch.load(os.path.join(
self.root, self.data_name, "degree_centrality"))
self.dataset.ori_degree = centrality1
min_val = torch.min(centrality1)
max_val = torch.max(centrality1)
centrality1 = (centrality1 - min_val) / (max_val - min_val)
centrality2 = torch.load(os.path.join(
self.root, self.data_name, "clustering_coefficients"))
min_val = torch.min(centrality2)
max_val = torch.max(centrality2)
centrality2 = (centrality2 - min_val) / (max_val - min_val)
self.dataset.degree_list = centrality1
self.dataset.cluster_list = centrality2
print("already has the required matrix")
return centrality
except:
print("there is no {} for {}".format(method, self.data_name))
G = nx.Graph()
G.add_nodes_from(range(self.dataset.data.num_node))
matrix = self.dataset.adj.tocsr()
rows, cols = matrix.nonzero()
G.add_edges_from(zip(rows, cols))
# G.add_edges_from(
# zip(self.dataset.data.edge.col.tolist(), self.dataset.data.edge.row.tolist()))
if method == "Betweenness_centrality":
centrality = nx.betweenness_centrality(G)
elif method == "Engienvector_centrality":
centrality = nx.eigenvector_centrality(G)
elif method == 'Closeness_centrality':
centrality = nx.closeness_centrality(G)
elif method == 'clustering_coefficients':
centrality = nx.clustering(G)
elif method == "pagerank":
centrality = nx.pagerank(G)
elif method == "degree_centrality":
centrality = nx.degree_centrality(G)
elif method == "together":
centrality = nx.degree_centrality(G)
centrality = torch.tensor(
[centrality[node] for node in range(self.dataset.data.num_node)])
torch.save(centrality, os.path.join(
self.root, self.data_name, "degree_centrality"))
centrality = nx.clustering(G)
centrality = torch.tensor(
[centrality[node] for node in range(self.dataset.data.num_node)])
torch.save(centrality, os.path.join(
self.root, self.data_name, "clustering_coefficients"))
centrality = nx.eigenvector_centrality(G)
centrality = torch.tensor(
[centrality[node] for node in range(self.dataset.data.num_node)])
torch.save(centrality, os.path.join(
self.root, self.data_name, "Engienvector_centrality"))
if method != "together":
centrality = torch.tensor(
[centrality[node] for node in range(self.dataset.data.num_node)])
torch.save(centrality, os.path.join(
self.root, self.data_name, method))
print("Spending {:.4f}s in finishing mining task".format(
time.time() - start_time))
min_val = torch.min(centrality)
max_val = torch.max(centrality)
centrality = (centrality - min_val) / (max_val - min_val)
self.dataset.mining_list = centrality
if method == "together":
centrality1 = torch.load(os.path.join(
self.root, self.data_name, "degree_centrality"))
centrality2 = torch.load(os.path.join(
self.root, self.data_name, "clustering_coefficients"))
try:
centrality3 = torch.load(os.path.join(
self.root, self.data_name, "Engienvector_centrality"))
except:
print("there is no engienvector for the current dataset!")
centrality3 = torch.load(os.path.join(
self.root, self.data_name, "degree_centrality"))
c = 0
for i, centrality in enumerate([centrality1, centrality2, centrality3]):
num_outliers = int(0.05 * centrality.size()[0])
print(num_outliers)
arr = centrality
_, indices = torch.topk(arr.abs(), num_outliers)
outliers = arr[indices]
arr[indices] = 1.0
min_value = arr.min()
max_value = arr[arr != 1.0].max()
centrality = (arr - min_value) / \
(max_value - min_value)
centrality[arr == 1.0] = 1.0
if i == 0:
centrality1 = centrality
if i == 1:
centrality2 = centrality
if i == 2:
centrality3 = centrality
centrality = a*centrality1 + b*centrality2 + c*centrality3
min_val = torch.min(centrality)
max_val = torch.max(centrality)
centrality = (centrality - min_val) / (max_val - min_val)
self.dataset.mining_list = centrality
print("already has the required matrix")
return centrality