-
Notifications
You must be signed in to change notification settings - Fork 0
/
generate_dataset.py
89 lines (63 loc) · 1.68 KB
/
generate_dataset.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
''' This file generates a dataset to be used and evaluated by the models '''
from __future__ import division
import numpy as np
import scipy as sp
import utils
import pickle
from itertools import *
from distributions import Probit
D = 32
period = 150
reps =100
# Generate a transition matrix
sigma = np.zeros((32, 32))
for i in xrange(D):
sigma[i, i] = 90
if i-1 >= 0:
sigma[i, i-1] = 70
else:
sigma[i,i] += 70
if i-2 >= 0:
sigma[i, i-2] = 50
else:
sigma[i,i] += 50
if i-3 >= 0:
sigma[i, i-3] = 30
else:
sigma[i,i] += 30
if i + 1 < D:
sigma[i, i+1] = 70
else:
sigma[i,i] += 70
if i + 2 < D:
sigma[i, i+2] = 50
else:
sigma[i,i] += 50
if i + 3 < D:
sigma[i, i+3] = 30
else:
sigma[i,i] += 30
# Normalize it
sigma /= sigma.sum(axis=1).astype(np.float)
current = np.random.randint(0, D)
seq = [current]
for i in xrange(period-1):
state = np.random.choice(range(D), p=sigma[current, :])
seq.append(state)
current = state
chain = chain(*repeat(seq, reps))
bin_vectors = [utils.bin_vec(state) for state in chain]
latent_states = np.matrix(bin_vectors)
latent_states = np.transpose(latent_states)
# W weight matrix
W = np.matrix([[5,2,2,0,0], [1,4,2,1, 0], [1,1,4,1,1], [0,1,2,4,2]])
# Normalize W
W = W / W.sum(axis=1).astype(np.float)
# Center it
W = W - W.mean(axis=1)
# Generate observations
obs = [Probit(W, latent_states[:, i]).rvs().tolist()[0] for i in xrange(latent_states.shape[1])]
observations = np.matrix(obs)
# Save it
with open('synt_data.pickle', 'w') as f:
pickle.dump({'states':latent_states, 'obs':observations, 'W':W}, f)