-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathviterbi.cpp
158 lines (139 loc) · 6.7 KB
/
viterbi.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
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
146
147
148
149
150
151
152
153
154
155
156
157
158
// @author: nabin
// conda install -c conda-forge gxx -> if older version of g++ is present
// g++ -fPIC -shared -o viterbi.so viterbi.cpp -O3
#include <iostream>
#include <vector>
#include <unordered_map>
#include <algorithm>
#include <limits>
#include <tuple>
#include <set>
bool success = false;
std::vector<int> exclude_states_in_c; //global exclude states
// Function prototype for viterbi_main with C linkage specification
extern "C" void run_viterbi(std::vector<int> observation, int num_observations, std::vector<int> states, int num_states, std::vector<std::vector<double>> transition_matrix, std::vector<std::vector<double>> emission_matrix, std::vector<double> initial_matrix, std::vector<int> states_to_work_python);
// Implementation of viterbi with C linkage specification
extern "C" void viterbi(std::vector<int> observation, int num_observations, std::vector<int> states, int num_states, std::vector<std::vector<double>> transition_matrix, std::vector<std::vector<double>> emission_matrix, std::vector<double> initial_matrix, std::vector<int> states_to_work_python)
{
// Run viterbi program
std::vector<std::vector<double>> trellis(num_observations, std::vector<double>(num_states, 0));
std::unordered_map<int, std::vector<int>> path; //an empty unordered_map
std::vector<int> states_to_work_with;
for (int x=0; x<states_to_work_python.size(); x++)
{
if (std::find(exclude_states_in_c.begin(), exclude_states_in_c.end(), states_to_work_python[x]) == exclude_states_in_c.end())
{
states_to_work_with.push_back(states_to_work_python[x]);
}
}
for (int state: states_to_work_with)
{
trellis[0][state] = initial_matrix[state] + emission_matrix[state][observation[0]];
path[state].push_back(state);
}
for (int observation_index=1; observation_index<num_observations; observation_index++)
{
std::unordered_map<int, std::vector<int>> new_path;
for (int state: states_to_work_with)
{
double max_prob = -std::numeric_limits<float>::infinity();
int possible_state = -1;
for (int previous_state: states_to_work_with)
{
auto it = std::find(path[previous_state].begin(), path[previous_state].end(), state);
if (it != path[previous_state].end())
{
continue;
}
double prob = trellis[observation_index - 1][previous_state] + transition_matrix[previous_state][state] + emission_matrix[state][observation[observation_index]];
if (prob > max_prob)
{
max_prob = prob;
possible_state = previous_state;
}
}
double probability = max_prob;
if (possible_state == -1)
{
if (success == true){return;}
auto max_it = std::max_element(states_to_work_with.begin(), states_to_work_with.end(),
[&](const auto& state1, const auto& state2) {
return trellis[observation_index - 1][state1] <
trellis[observation_index - 1][state2];
});
auto probability = trellis[observation_index - 1][*max_it];
auto state = *max_it;
exclude_states_in_c.insert(exclude_states_in_c.end(), path[state].begin(), path[state].end());
std::vector<int> sub_observation(observation.begin() + observation_index, observation.end());
run_viterbi(sub_observation, sub_observation.size(), states, num_states, transition_matrix, emission_matrix, initial_matrix, states_to_work_python);
}
trellis[observation_index][state] = probability;
new_path[state] = path[possible_state];
new_path[state].push_back(state);
}
path = new_path;
}
auto max_it = std::max_element(states_to_work_with.begin(), states_to_work_with.end(),
[&](const auto& state1, const auto& state2) {
return trellis[num_observations - 1][state1] <
trellis[num_observations - 1][state2];
});
auto probability = trellis[num_observations - 1][*max_it];
auto state = *max_it;
exclude_states_in_c.insert(exclude_states_in_c.end(), path[state].begin(), path[state].end());
success = true;
return;
}
extern "C" void run_viterbi(std::vector<int> observation, int num_observations, std::vector<int> states, int num_states, std::vector<std::vector<double>> transition_matrix, std::vector<std::vector<double>> emission_matrix, std::vector<double> initial_matrix, std::vector<int> states_to_work_python)
{
viterbi(observation, num_observations, states, num_states, transition_matrix, emission_matrix, initial_matrix, states_to_work_python);
return;
}
// Implementation of viterbi_main with C linkage specification
extern "C" int* viterbi_main(int* obs, int num_observations, int num_states, double* transt, double* emiss, double* init, int* exclude_s, int exclude_s_len)
{
success = false;
std::vector<int> observations(num_observations);
std::vector<int> states(num_states);
std::vector<std::vector<double>> transition_matrix(num_states, std::vector<double>(num_states));
std::vector<std::vector<double>> emission_matrix(num_states, std::vector<double>(20));
std::vector<double> initial_matrix(num_states);
std::vector<int> exclude_stat(exclude_s_len);
for (int i = 0; i<num_states; i++)
{
initial_matrix[i] = init[i];
states[i] = i;
for (int j = 0; j<num_states; j++)
{
transition_matrix[i][j] = transt[i*num_states + j];
}
}
for (int i = 0; i<num_states; i++)
{
for (int j = 0; j<20; j++)
{
emission_matrix[i][j] = emiss[i*20 + j];
}
}
for (int i = 0; i<exclude_s_len; i++)
{
exclude_stat[i] = exclude_s[i];
}
for (int i = 0; i<num_observations; i++)
{
observations[i] = obs[i];
}
std::vector<int> states_to_work_python;
for (int x=0; x<num_states; x++)
{
if (std::find(exclude_stat.begin(), exclude_stat.end(), states[x]) == exclude_stat.end())
{
states_to_work_python.push_back(states[x]);
}
}
exclude_states_in_c.clear();
run_viterbi(observations, num_observations, states, num_states, transition_matrix, emission_matrix, initial_matrix, states_to_work_python);
std::set<int> s(exclude_states_in_c.begin(), exclude_states_in_c.end());
int * result = exclude_states_in_c.data();
return result;
}