-
Notifications
You must be signed in to change notification settings - Fork 0
/
convergence.py
145 lines (119 loc) · 4.57 KB
/
convergence.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
__author__ = 'shreyarajpal'
from time import time
import basics
from random import choice, seed
from GibbsSampling import getRandomSample
def GibbsSamplerConvergence(w1, w2, actualWord1, actualWord2):
global factorLookUp
n1 = len(w1)
n2 = len(w2)
#Create a factor look-up list
sk1, sk2, ps = basics.findingSkips(w1, w2)
factorLookUp = basics.getFactor(n1, n2, sk1, sk2, ps)
#Generate initial assignment
seed(time())
assignment = [choice(basics.characterArray) for i in xrange(n1 + n2)]
samples = [{t:0 for t in basics.characterArray} for i in xrange(n1 + n2)]
burnInSamples = [{t:0 for t in basics.characterArray} for i in xrange(n1 + n2)]
prevLogLikelihoodOfMLA = 500
burninInfo = []
count = 0
while(True):
flag = False
for j in xrange(n1 + n2):
sample = getRandomSample(j, assignment, w1, w2)
assignment[j] = sample
#burnInSamples[j][sample] += 1
for x in xrange(len(assignment)):
burnInSamples[x][assignment[x]] += 1
count += 1
if count>100:
logLikelihoodOfMLA = 0
for k in xrange(n1 + n2):
logLikelihoodOfMLA += float(max(burnInSamples[k].values()))/count
if abs(logLikelihoodOfMLA - prevLogLikelihoodOfMLA)<0.0002:
flag = True
break
else:
prevLogLikelihoodOfMLA = logLikelihoodOfMLA
burninInfo.append(logLikelihoodOfMLA)
if flag:
break
iterationInfo = []
count = 0
for i in xrange(10000):
for j in xrange(n1 + n2):
count += 1
sample = getRandomSample(j, assignment, w1, w2)
assignment[j] = sample
if count%10==0:
for x in xrange(n1 + n2):
samples[x][assignment[x]] += 1
MLA = ''
for k in samples:
maxMarginal = max(k.values())
for l in k.keys():
if k[l]==maxMarginal:
MLA += l
break
word1 = MLA[:len(w1)]
word2 = MLA[len(w1):]
iterationInfo.append((word1==actualWord1, word2==actualWord2))
if count >= 20000:
break
return iterationInfo
def GibbsSamplerRandomisedConvergence(w1, w2, actualWord1, actualWord2):
global factorLookUp
n1 = len(w1)
n2 = len(w2)
#Create a factor look-up list
sk1, sk2, ps = basics.findingSkips(w1, w2)
factorLookUp = basics.getFactor(n1, n2, sk1, sk2, ps)
#Generate initial assignment
seed(time())
assignment = [choice(basics.characterArray) for i in xrange(n1 + n2)]
samples = [{t:0 for t in basics.characterArray} for i in xrange(n1 + n2)]
burnInSamples = [{t:0 for t in basics.characterArray} for i in xrange(n1 + n2)]
prevLogLikelihoodOfMLA = 500
burninInfo = []
count = 0
while(True):
varSampled = choice(xrange(n1 + n2))
sample = getRandomSample(varSampled, assignment, w1, w2)
assignment[varSampled] = sample
#burnInSamples[varSampled][sample] += 1
for x in xrange(len(assignment)):
burnInSamples[x][assignment[x]] += 1
count += 1
if count>100:
logLikelihoodOfMLA = 0
for j in xrange(n1 + n2):
logLikelihoodOfMLA += float(max(burnInSamples[j].values()))/count
if abs(logLikelihoodOfMLA - prevLogLikelihoodOfMLA)<0.0002:
break
else:
prevLogLikelihoodOfMLA = logLikelihoodOfMLA
burninInfo.append(logLikelihoodOfMLA)
#print 'Burn-in ', len(burninInfo)
iterationInfo = []
count = 0
for i in xrange(20000):
varSampled = choice(xrange(n1 + n2))
count += 1
sample = getRandomSample(varSampled, assignment, w1, w2)
assignment[varSampled] = sample
if count%10==0:
for x in xrange(n1 + n2):
samples[x][assignment[x]] += 1
MLA = ''
for k in samples:
maxMarginal = max(k.values())
for j in k.keys():
if k[j]==maxMarginal:
MLA += j
break
word1 = MLA[:len(w1)]
word2 = MLA[len(w1):]
iterationInfo.append((word1==actualWord1, word2==actualWord2))
return iterationInfo
#print GibbsSamplerRandomisedConvergence([542,949,830], [742,981,543,625,830,758], 'ade', 'atoner')