-
Notifications
You must be signed in to change notification settings - Fork 4
/
likelihood_computer.cpp
111 lines (97 loc) · 4.79 KB
/
likelihood_computer.cpp
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
#include "hdp.h"
#include "likelihood_computer.h"
#include "utils.h"
#include "corpus.h"
#include "state.h"
#include <map>
#include <iostream>
#include <vector>
using namespace std;
double likelihood_computer::compute_harmonic_mean_predictive_likelihood(const hdp* const * posterior_samplers,
const counts* const* before_update_counts,
int m_posterior_samplers)
{
double likelihood = 0;
if (posterior_samplers == NULL) {
return likelihood;
}
int m_size_vocab = posterior_samplers[0]->get_size_vocab();
double m_eta = posterior_samplers[0]->get_eta();
vector<double> log_likelihood;
log_likelihood.assign(m_posterior_samplers, 0);
for (int posterior_sampler_id = 0; posterior_sampler_id < m_posterior_samplers; ++posterior_sampler_id) {
int m_old_topics = before_update_counts[posterior_sampler_id]->m_word_counts_by_z.size();
int m_new_topics = posterior_samplers[posterior_sampler_id]->get_topics_number();
for (int k = 0; k < m_old_topics; ++k) {
log_likelihood[posterior_sampler_id] += lgamma(m_size_vocab * m_eta + before_update_counts[posterior_sampler_id]->m_word_counts_by_z[k]) -
lgamma(m_size_vocab * m_eta + posterior_samplers[posterior_sampler_id]->get_word_topic_counts(k));
for (int w = 0; w < m_size_vocab; ++w) {
int cur_word_topic_count = posterior_samplers[posterior_sampler_id]->get_word_topic_counts(k, w);
int cur_word_topic_count_before_update = before_update_counts[posterior_sampler_id]->m_word_counts_by_zw[k][w];
if (cur_word_topic_count != cur_word_topic_count_before_update) {
log_likelihood[posterior_sampler_id] += lgamma(cur_word_topic_count + m_eta) -
lgamma(cur_word_topic_count_before_update + m_eta);
}
}
}
for (int k = m_old_topics; k < m_new_topics; ++k) {
log_likelihood[posterior_sampler_id] += lgamma(m_size_vocab * m_eta) -
lgamma(m_size_vocab * m_eta +
posterior_samplers[posterior_sampler_id]->get_word_topic_counts(k));
for (int w = 0; w < m_size_vocab; ++w) {
if (posterior_samplers[posterior_sampler_id]->get_word_topic_counts(k, w) > 0) {
log_likelihood[posterior_sampler_id] +=
lgamma(posterior_samplers[posterior_sampler_id]->get_word_topic_counts(k, w) + m_eta) -
lgamma(m_eta);
}
}
}
}
invert_element_sign(log_likelihood);
likelihood = -log_sum_exp_trick(log_likelihood) + log(m_posterior_samplers);
return likelihood;
}
double likelihood_computer::compute_harmonic_mean_predictive_likelihood(const hdp_dynamic* const * posterior_samplers,
const counts* const* before_update_counts,
int m_posterior_samplers)
{
double likelihood = 0;
if (posterior_samplers == NULL) {
return likelihood;
}
int m_size_vocab = posterior_samplers[0]->get_size_vocab();
double m_eta = posterior_samplers[0]->get_eta();
vector<double> log_likelihood;
log_likelihood.assign(m_posterior_samplers, 0);
for (int posterior_sampler_id = 0; posterior_sampler_id < m_posterior_samplers; ++posterior_sampler_id) {
int m_old_topics = before_update_counts[posterior_sampler_id]->m_word_counts_by_z.size();
int m_new_topics = posterior_samplers[posterior_sampler_id]->get_topics_number();
for (int k = 0; k < m_old_topics; ++k) {
log_likelihood[posterior_sampler_id] += lgamma(m_size_vocab * m_eta + before_update_counts[posterior_sampler_id]->m_word_counts_by_z[k]) -
lgamma(m_size_vocab * m_eta + posterior_samplers[posterior_sampler_id]->get_word_topic_counts(k));
for (int w = 0; w < m_size_vocab; ++w) {
int cur_word_topic_count = posterior_samplers[posterior_sampler_id]->get_word_topic_counts(k, w);
int cur_word_topic_count_before_update = before_update_counts[posterior_sampler_id]->m_word_counts_by_zw[k][w];
if (cur_word_topic_count != cur_word_topic_count_before_update) {
log_likelihood[posterior_sampler_id] += lgamma(cur_word_topic_count + m_eta) -
lgamma(cur_word_topic_count_before_update + m_eta);
}
}
}
for (int k = m_old_topics; k < m_new_topics; ++k) {
log_likelihood[posterior_sampler_id] += lgamma(m_size_vocab * m_eta) -
lgamma(m_size_vocab * m_eta +
posterior_samplers[posterior_sampler_id]->get_word_topic_counts(k));
for (int w = 0; w < m_size_vocab; ++w) {
if (posterior_samplers[posterior_sampler_id]->get_word_topic_counts(k, w) > 0) {
log_likelihood[posterior_sampler_id] +=
lgamma(posterior_samplers[posterior_sampler_id]->get_word_topic_counts(k, w) + m_eta) -
lgamma(m_eta);
}
}
}
}
invert_element_sign(log_likelihood);
likelihood = -log_sum_exp_trick(log_likelihood) + log(m_posterior_samplers);
return likelihood;
}