-
Notifications
You must be signed in to change notification settings - Fork 0
/
experiment.py
146 lines (128 loc) · 5.29 KB
/
experiment.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
import numpy as np
import matplotlib.pyplot as plt
from scipy import sparse
import time
from multiprocessing import Pool
from em import initialize, em, mle_em, mle_em_with_obj, mle_gd_with_obj
def run_gd(arg):
np.random.seed()
max_step, threshold, sparse, ix, norm = arg
start = time.time()
y_rand = np.random.poisson(Ax)
# x_initial = np.random.randn(len(x_flat))
alpha = 1.
if norm > 1:
alpha = 0.1
if sparse:
x_init = np.linalg.inv((A.T@A).todense())@(A.T@y_rand)
else:
x_init = np.linalg.inv((A.T@A))@(A.T@y_rand)
x_init = np.array(x_init).flatten()
x_init = np.array([v if v>0.1 else 10 for v in x_init])
print(f"process: {ix: 2d}, y[:10]: {y_rand[:10]}")
x_et, diff, mse, objs, step = mle_gd_with_obj(max_step, A, y_rand, x_true=x_flat, threshold=threshold, x_initial=x_init, sparse=sparse, alpha=alpha)
print(f"process: {ix: 2d} finished. step: {step: 2d}, mse: {mse: 8.2f}, diff: {diff: 8.2f} time consuming: {time.time() - start: 8.1f} seconds")
return x_et, diff, mse, step
def run(arg):
np.random.seed()
max_step, threshold, sparse, ix = arg
start = time.time()
y_rand = np.random.poisson(Ax)
alpha = 1
# x_initial = np.random.randn(len(x_flat))
print(f"process: {ix: 2d}, y[:10]: {y_rand[:10]}")
x_et, diff, mse, objs, step = mle_em_with_obj(max_step, A, y_rand, x_true=x_flat, threshold=threshold, x_initial=None, sparse=sparse)
print(f"process: {ix: 2d} finished. step: {step: 2d}, mse: {mse: 8.2f}, diff: {diff: 8.2f} time consuming: {time.time() - start: 8.1f} seconds")
return x_et, diff, mse, step
def mc_mp(n, max_step=2000, threshold=10., sparse=True):
pool = Pool()
args = [(max_step, threshold, sparse, i) for i in range(n)]
result = pool.map(run, args)
pool.close()
pool.join()
return result
def mc_mp_gd(n, max_step=10000, threshold=10., sparse=True, norm=1):
pool = Pool()
args = [(max_step, threshold, sparse, i, norm) for i in range(n)]
result = pool.map(run_gd, args)
pool.close()
pool.join()
return result
def mc_run(n, max_step=500, threshold=10, sparse=True):
etms = []
for i in range(n):
print(f"{i}th run start =>")
y = np.random.poisson(Ax)
x_et, diff, mse, step = mle_em(max_step, A, y, x_true=x_flat, threshold=threshold, x_initial=None, sparse=sparse)
etms.append([x_et, diff, mse, step])
return etms
def multi_mc(A_, x_):
global A
global Ax
norms = np.logspace(-1, 1, 11)
pows = np.linspace(-1, 1, 11)
for norm, pow in zip(norms, pows):
print(f"norm of A is: {norm: .4f}")
A = A_original*norm
Ax = A @ x_flat
for repeat in range(200//n_para):
try:
etms = mc_mp(n=n_para, threshold=e_stop)
etms_np = np.array(etms, dtype=object)
np.save(f"{home}/etms_{time.time()}_batch_{n_para}_stop_{e_stop}_norm_10e{pow: .2f}.npy", etms_np)
except Exception as e:
print(f"except at {repeat}-th multiprocessing: {e}")
def multi_mc_gd(A_, x_):
global A
global Ax
norms = np.logspace(-1, 1, 11)
pows = np.linspace(-1, 1, 11)
for norm, pow in zip(norms, pows):
print(f"norm of A is: {norm: .4f}")
A = A_original*norm
Ax = A @ x_flat
for repeat in range(epoch//n_para):
try:
etms = mc_mp_gd(n=n_para, threshold=e_stop, norm=norm)
etms_np = np.array(etms, dtype=object)
np.save(f"{home}/etms_{time.time()}_batch_{n_para}_stop_{e_stop}_norm_10e{pow: .2f}.npy", etms_np)
except Exception as e:
print(f"except at {repeat}-th multiprocessing: {e}")
if __name__ == "__main__":
# A_original = sparse.load_npz("data/simulated_large_A_117_100.npz")
# x_flat = np.load("data/simulated_large_x_117_100.npy")
# A_original = sparse.load_npz("data/simulated_large_A_56_50.npz")
# x_flat = np.load("data/simulated_large_x_56_50.npy")
A_original = sparse.load_npz("data/simulated_large_A_23_10.npz")
x_flat = np.load("data/simulated_large_x_23_10.npy")
# y = np.load("constant_y/y.npy")
# y = np.random.poisson(Ax)
print("Image vector x is of shape {}".format(np.shape(x_flat)))
print("Mixing matrix A is of shape {}".format(np.shape(A_original)))
# print("Observation matrix Y is of shape {}".format(np.shape(y)))
x_flat[0] = 20
x_flat[90] =20
epoch = 100
n_para = 100
e_stop = 0.0001
home = "et10_gd_rerun"
# multi_mc(A_original, x_flat)
# multi_mc(A_original, x_flat)
multi_mc_gd(A_original, x_flat)
# nm = 1
# A = A_original*nm
# Ax = A @ x_flat
# run((10000, e_stop, True, 0))
# run_gd((10000, e_stop, True, 0, nm))
# etms = mc_run(2)
# norm = 1
# A = A_original*norm
# Ax = A @ x_flat
# for repeat in range(200//n_para):
# try:
# etms = mc_mp(n=n_para, threshold=e_stop)
# etms_np = np.array(etms, dtype=object)
# np.save(f"et50/etms_{time.time()}_batch_{n_para}_stop_{e_stop}_norm_{norm}.npy", etms_np)
# # np.save(f"constant_y/etms_{time.time()}_{n_para}_{e_stop}.npy", etms_np)
# except Exception as e:
# print(f"except at {repeat}-th multiprocessing: {e}")