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
|
import numpy as np
import torch
import pickle
import time
from itertools import combinations
from scipy.special import comb
from multiprocessing import Pool
def schuffle(W, x):
n = W.shape[0]
idx=np.random.permutation(n)
W_new=W[idx,:][:,idx]
x_new=x[idx]
return W_new , x_new
def block_model_adj(com, p_r, q):
n=len(com)
W=np.zeros((n, n), dtype=np.int64)
for i in range(n):
for j in range(i+1,n):
if com[i] == com[j]:
prob = p_r[com[i]]
else:
prob = q
if np.random.binomial(1, prob) == 1:
W[i,j] = 1
W[j,i] = 1
return W
def block_model(sel_r, clust_size_min, clust_size_max, p_r, q):
com = []
for r in sel_r:
clust_size = np.random.randint(clust_size_min, clust_size_max, size=1)[0]
com.append(np.repeat(r, clust_size, axis=0))
com = np.concatenate(com)
W = block_model_adj(com, p_r, q)
return W, com
def adj_to_edge_list(W):
n = W.shape[0]
e = int(np.sum(W))
edge_index = np.zeros((2, e), dtype=np.int64)
counter = 0
for i in range(n):
for j in range(n):
if W[i,j] == 1:
edge_index[0, counter] = i
edge_index[1, counter] = j
counter += 1
assert counter == e
return edge_index
SMB_ENVIRONMENT_CONFIG = {
'r1': 4,
'r2': 3,
's1': 3,
's2': 2,
'p_r': np.array([0.9, 0.8, 0.7, 0.6, 0.5, 0.4, 0.3]),
'size_min': 20,
'size_max': 40,
'q': 0.1,
'ntypes': 8,
}
DOMAIN_SELECTION_DICT = dict(zip([frozenset(sel_r1) for sel_r1 in combinations(range(SMB_ENVIRONMENT_CONFIG['r1']), SMB_ENVIRONMENT_CONFIG['s1'])], range(comb(SMB_ENVIRONMENT_CONFIG['r1'], SMB_ENVIRONMENT_CONFIG['s1'], exact=True))))
CLASS_SELECTION_DICT = dict(zip([frozenset(sel_r2) for sel_r2 in combinations(range(SMB_ENVIRONMENT_CONFIG['r2']), SMB_ENVIRONMENT_CONFIG['s2'])], range(comb(SMB_ENVIRONMENT_CONFIG['r2'], SMB_ENVIRONMENT_CONFIG['s2'], exact=True))))
def encode_domain_selection(sel_r1):
return DOMAIN_SELECTION_DICT[frozenset(sel_r1)]
def encode_class_selection(sel_r2):
return CLASS_SELECTION_DICT[frozenset(sel_r2)]
class SBMEnvironmentGraph():
def __init__(self, config=SMB_ENVIRONMENT_CONFIG):
# parameters
self.r1 = SMB_ENVIRONMENT_CONFIG['r1']
self.r2 = SMB_ENVIRONMENT_CONFIG['r2']
self.s1 = SMB_ENVIRONMENT_CONFIG['s1']
self.s2 = SMB_ENVIRONMENT_CONFIG['s2']
assert self.s1 > 0 and self.s1 < self.r1
assert self.s2 > 0 and self.s2 < self.r2
self.p_r = SMB_ENVIRONMENT_CONFIG['p_r']
self.p_r1 = self.p_r[0::2]
self.p_r2 = self.p_r[1::2]
assert len(self.p_r1) == self.r1
assert len(self.p_r2) == self.r2
self.size_min = SMB_ENVIRONMENT_CONFIG['size_min']
self.size_max = SMB_ENVIRONMENT_CONFIG['size_max']
self.q = SMB_ENVIRONMENT_CONFIG['q']
self.ntypes = SMB_ENVIRONMENT_CONFIG['ntypes']
# data
self.x = None
self.y = None
self.edge_index = None
self.domain_id = None
# generate
self.generate()
def validate(self):
assert isinstance(self.x, torch.Tensor)
assert isinstance(self.y, torch.Tensor)
assert isinstance(self.edge_index, torch.Tensor)
assert isinstance(self.domain_id, np.ndarray)
assert self.x.ndim == 1
assert self.y.ndim == 0
assert self.edge_index.ndim == 2 and self.edge_index.size(0) == 2
assert self.domain_id.ndim == 0
assert self.x.dtype == torch.int64
assert self.y.dtype == torch.int64
assert self.edge_index.dtype == torch.int64
assert self.domain_id.dtype == np.int64
def get(self):
self.validate()
graph_data = {
'x': self.x,
'y': self.y,
'edge_index': self.edge_index,
}
return graph_data, self.domain_id
def stat(self):
self.validate()
stat = {
'# of nodes': self.x.size(0),
'# of edges': self.edge_index.size(1)//2,
'avg degree': self.edge_index.size(1)/self.x.size(0)/2,
}
return stat
def generate(self):
# select environment (domain defining) communities
sel_r1 = np.random.choice(self.r1, self.s1, replace=False)
domain_id = encode_domain_selection(sel_r1)
# select the target (class defining) communities
sel_r2 = np.random.choice(self.r2, self.s2, replace=False)
y = encode_class_selection(sel_r2)
# combine the sel and edge prob lists
sel_r2 = sel_r2 + self.r1
sel_r = np.concatenate([sel_r1, sel_r2])
p_r = np.concatenate([self.p_r1, self.p_r2])
# generate adj and com list
W, com = block_model(sel_r, self.size_min, self.size_max, self.p_r, self.q)
# generate iid random node features
x = np.random.randint(self.ntypes, size=W.shape[0])
# shuffle
W, x = schuffle(W, x)
# conversion
self.x = torch.from_numpy(x).long()
self.y = torch.tensor(y).long()
self.edge_index = torch.from_numpy(adj_to_edge_list(W)).long()
self.domain_id = np.array(domain_id, dtype=np.int64)
def generate_func(i):
return SBMEnvironmentGraph()
def generate_sbm_graph_list(n_samples, n_processes=1):
np.random.seed(0)
start_time = time.time()
sbm_graph_list = []
with Pool(processes=n_processes) as pool:
for i, sbmg in enumerate(pool.imap_unordered(generate_func, range(n_samples))):
sbm_graph_list.append(sbmg)
print(f'Progress: {i/n_samples*100:.2f}%', end='\r', flush=True)
print(' '*50, end='\r')
print(f'Total generation time: {time.time() - start_time:.2f} secs.')
return sbm_graph_list
|