-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathgen_simulation_data.py
131 lines (102 loc) · 3.43 KB
/
gen_simulation_data.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
import pickle
import argparse
import os
import numpy as np
def get_args():
parser = argparse.ArgumentParser()
parser.add_argument('--type', type=str, default='special',
choices=['origin', 'special'])
parser.add_argument('--gmm-gen-kk', type=int, default=4)
parser.add_argument('--gmm-gen-std', type=float, default=2)
parser.add_argument('--gmm-gen-n', type=int, default=2000)
parser.add_argument('--random-seed', type=int, default=23498353)
parser.add_argument('--save-dir', type=str, default='./data/GMMs')
parser.add_argument('--save-name', type=str, default='gmm-2d-syn-set.pkl')
return parser.parse_args()
def gen_data_special(kk, n, prior_std):
np.random.seed(23498353)
# dx1 = np.random.rand() * 3
# dx2 = np.random.rand() * 3
dx1 = prior_std[0] ** 2
dx2 = prior_std[1] ** 2
coef = np.random.rand() * np.sqrt(dx1) * np.sqrt(dx2)
cc = np.array([[dx1, coef], [coef, dx2]])
mu = np.random.multivariate_normal([0,0], cc, size=kk)
# print(mu.shape)
# print(mu)
mu[1] += [-1,2.5]
mu[3] += [1,-0]
cov = []
for k in range(kk):
if k==0:
dx1 = 1
dx2 = 0.05
cov.append( np.array([[dx1, 0], [0, dx2]]) )
elif k==2:
dx1 = 0.05
dx2 = 1
cov.append( np.array([[dx1, 0], [0, dx2]]) )
else:
dx1 = 1
dx2 = 1
cov.append( np.array([[dx1, 0], [0, dx2]]) )
x, y = [], []
for i in range(n):
k = np.random.randint(kk)
x.append( np.random.multivariate_normal(mu[k], cov[k]) )
y.append(k)
x = np.array(x).astype(np.float32)
y = np.array(y).astype(np.int)
return x, y
def gen_data_origin(kk, n, prior_std):
# dx1 = np.random.rand() * 3
# dx2 = np.random.rand() * 3
dx1 = prior_std[0] ** 2
dx2 = prior_std[1] ** 2
coef = np.random.rand() * np.sqrt(dx1) * np.sqrt(dx2)
cc = np.array([[dx1, coef], [coef, dx2]])
mu = np.random.multivariate_normal([0,0], cc, size=kk)
cov = []
for k in range(kk):
# dx1 = np.random.rand() * (prior_std[0]**2) * 0.2
# dx2 = np.random.rand() * (prior_std[1]**2) * 0.2
dx1 = 1
dx2 = 1
# coef = np.random.rand() * np.sqrt(dx1) * np.sqrt(dx2)
coef = 0
cov.append( np.array([[dx1, coef], [coef, dx2]]) )
x, y = [], []
for i in range(n):
k = np.random.randint(kk)
x.append( np.random.multivariate_normal(mu[k], cov[k]) )
y.append(k)
x = np.array(x).astype(np.float32)
y = np.array(y).astype(np.int)
return x, y
def main():
args = get_args()
''' fix random seed '''
# SEED = 19260817
# # NP_SEED = 10**9 + 9
# NP_SEED = 31894921
np.random.seed(args.random_seed)
if not os.path.exists(args.save_dir):
os.makedirs(args.save_dir)
gen_data_params = {
'kk': args.gmm_gen_kk,
'n': args.gmm_gen_n,
'prior_std': [args.gmm_gen_std, args.gmm_gen_std],
}
if args.type == 'origin':
x, y = gen_data_origin(**gen_data_params)
elif args.type == 'special':
x, y = gen_data_special(**gen_data_params)
else:
raise ValueError
save_path = '{}/{}'.format(args.save_dir, args.save_name)
with open(save_path, 'wb') as f:
pickle.dump({'x':x, 'y':y}, f)
print('the generated dataset has been saved to `{}`'.format(save_path))
return
if __name__ == '__main__':
main()