-
Notifications
You must be signed in to change notification settings - Fork 1
/
metrics.py
141 lines (122 loc) · 7.24 KB
/
metrics.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
import numpy as np
import matplotlib.pyplot as plt
from networkx import nx
import collections
from scipy.io import savemat
import time
import sys
from utils_network import *
# In the context of machine learning, KL(P|Q) is often called the information gain achieved if Q is used instead of P
def kl(p, q):
p = np.asarray(p, dtype=np.float)
q = np.asarray(q, dtype=np.float)
return np.sum(p * np.log(p / q))
def bd(p,q):
return -np.log(np.sum(np.sqrt(np.multiply(p, q))))
# --- the KL divergence of the degree of graphs --- #
def degree(output_dir):
print('Start computing graph degree')
random_G, random_bara_G, generated_network, original_network, GAE_network, NetGAN_network = load_data(output_dir)
ds = np.array(sorted([d for n, d in original_network.degree()], reverse=True))
mean = np.mean(ds)
max = np.max(ds)
original_network_ds = degree_distribution(original_network, mean, max)
random_G_ds = degree_distribution(random_G, mean, max)
random_bara_G_ds = degree_distribution(random_bara_G, mean, max)
generated_network_ds = degree_distribution(generated_network, mean, max)
GAE_network_ds = degree_distribution(GAE_network, mean, max)
# NetGAN_network_ds = degree_distribution(NetGAN_network, mean, max)
data_1 = kl(original_network_ds, generated_network_ds)
data_2 = kl(original_network_ds, random_G_ds)
data_3 = kl(original_network_ds, random_bara_G_ds)
data_4 = kl(original_network_ds, GAE_network_ds)
# data_5 = kl(original_network_ds, NetGAN_network_ds)
print('KL divergence between original network and the network generated by Music_GAN: {}'.format(data_1))
print('KL divergence between original network and the network generated by Random E-R: {}'.format(data_2))
print('KL divergence between original network and the network generated by Random B-A: {}'.format(data_3))
print('KL divergence between original network and the network generated by GAE: {}'.format(data_4))
# print('KL divergence between original network and the network generated by NETGAN: {}'.format(data_5))
# data_1 = bd(original_network_ds, generated_network_ds)
# data_2 = bd(original_network_ds, random_G_ds)
# data_3 = bd(original_network_ds, random_bara_G_ds)
# data_4 = bd(original_network_ds, GAE_network_ds)
# data_5 = bd(original_network_ds, NetGAN_network_ds)
# print('Bhattacharyya distance between original network and the network generated by Music_GAN: {}'.format(data_1))
# print('Bhattacharyya distance between original network and the network generated by Random E-R: {}'.format(data_2))
# print('Bhattacharyya distance between original network and the network generated by Random B-A: {}'.format(data_3))
# print('Bhattacharyya distance between original network and the network generated by GAE: {}'.format(data_4))
# print('Bhattacharyya distance between original network and the network generated by NETGAN: {}'.format(data_5))
def softmax(x, mean, max):
e_x = np.exp((x-max)/mean)
return e_x / e_x.sum()
def degree_distribution(graph, mean, max):
ds = np.array(sorted([d for n, d in graph.degree()], reverse=True))
ds[ds==0]=1
# ds = np.array([d for n, d in graph.degree()])
# degreeCount = collections.Counter(ds)
# deg, cnt = zip(*degreeCount.items())
return softmax(ds, mean, max)
# --- the KL divergence of the coefficient of graphs --- #
def coefficient(output_dir):
random_G, random_bara_G, generated_network, original_network, GAE_network, NetGAN_network = load_data(output_dir)
print('Start computing graph coefficient')
a = np.array(sorted([x[1] for x in nx.clustering(original_network).items()]))
mean = np.mean(a)
max = np.max(a)
original_network_coef = coef(original_network, mean, max)
random_G_coef = coef(random_G, mean, max)
random_bara_G_coef = coef(random_bara_G, mean, max)
generated_network_coef = coef(generated_network, mean, max)
GAE_network_coef = coef(GAE_network, mean, max)
# NetGAN_network_coef = coef(NetGAN_network, mean, max)
data_1 = kl(generated_network_coef, original_network_coef)
data_2 = kl(original_network_coef, random_G_coef)
data_3 = kl(original_network_coef, random_bara_G_coef)
data_4 = kl(GAE_network_coef, original_network_coef)
# data_5 = kl(GAE_network_coef, NetGAN_network_coef)
print('KL divergence between original network and the network generated by Music_GAN: {}'.format(data_1))
print('KL divergence between original network and the network generated by Random E-R: {}'.format(data_2))
print('KL divergence between original network and the network generated by Random B-A: {}'.format(data_3))
print('KL divergence between original network and the network generated by GAE: {}'.format(data_4))
# print('KL divergence between original network and the network generated by NETGAN: {}'.format(data_5))
# data_1 = bd(generated_network_coef, original_network_coef)
# data_2 = bd(original_network_coef, random_G_coef)
# data_3 = bd(original_network_coef, random_bara_G_coef)
# data_4 = bd(GAE_network_coef, original_network_coef)
# data_5 = bd(GAE_network_coef, NetGAN_network_coef)
# print('Bhattacharyya distance between original network and the network generated by Music_GAN: {}'.format(data_1))
# print('Bhattacharyya distance between original network and the network generated by Random E-R: {}'.format(data_2))
# print('Bhattacharyya distance between original network and the network generated by Random B-A: {}'.format(data_3))
# print('Bhattacharyya distance between original network and the network generated by GAE: {}'.format(data_4))
# print('Bhattacharyya distance between original network and the network generated by NETGAN: {}'.format(data_5))
def coef(graph, mean, max):
coef = np.array(sorted([x[1] for x in nx.clustering(graph).items()]))
return softmax(coef, mean, max)
def load_data(output_dir):
NetGAN_network = None
generated_network = np.load('{}/output_network.npy'.format(output_dir))
original_network = np.load('{}/org_network.npy'.format(output_dir))
GAE_network = np.load('{}/GAE_network.npy'.format(output_dir))
# NetGAN_network = np.load('{}/netgan_network.npy'.format(output_dir))[1]
n = original_network.shape[0]
generated_network = generated_network + np.identity(n)
original_network = original_network + np.identity(n)
original_network[original_network > 2] = 1
generated_network[generated_network > 2] = 1
GAE_network = GAE_network + np.identity(n)
m = int(np.sum(original_network))
random_G = nx.gnm_random_graph(n, m)
random_bara_G = nx.generators.random_graphs.barabasi_albert_graph(n, 800)
GAE_network = nx.from_numpy_matrix(GAE_network)
generated_network = nx.from_numpy_matrix(generated_network)
original_network = nx.from_numpy_matrix(original_network)
# if NetGAN_network.shape[0] < n:
# Net = np.zeros((n, n))
# Net[:NetGAN_network.shape[0], :NetGAN_network.shape[1]] = NetGAN_network
# NetGAN_network = Net
# NetGAN_network = nx.from_numpy_matrix(NetGAN_network[:n, :n])
return random_G, random_bara_G, generated_network, original_network, GAE_network, NetGAN_network
if __name__ == '__main__':
filename = 'coarsegraph_facebook.mat'
degree(filename)
coefficient(filename)