opener-opinion-detector-basic 2.0.0 → 2.0.1
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/README.md +2 -0
- data/ext/hack/Rakefile +0 -2
- data/lib/opener/opinion_detector_basic/version.rb +1 -1
- data/opener-opinion-detector-basic.gemspec +0 -1
- data/task/compile.rake +1 -1
- data/task/requirements.rake +0 -1
- metadata +2 -142
- data/core/vendor/src/crfsuite/AUTHORS +0 -1
- data/core/vendor/src/crfsuite/COPYING +0 -27
- data/core/vendor/src/crfsuite/ChangeLog +0 -103
- data/core/vendor/src/crfsuite/INSTALL +0 -236
- data/core/vendor/src/crfsuite/Makefile.am +0 -19
- data/core/vendor/src/crfsuite/Makefile.in +0 -783
- data/core/vendor/src/crfsuite/README +0 -183
- data/core/vendor/src/crfsuite/aclocal.m4 +0 -9018
- data/core/vendor/src/crfsuite/autogen.sh +0 -38
- data/core/vendor/src/crfsuite/compile +0 -143
- data/core/vendor/src/crfsuite/config.guess +0 -1502
- data/core/vendor/src/crfsuite/config.h.in +0 -198
- data/core/vendor/src/crfsuite/config.sub +0 -1714
- data/core/vendor/src/crfsuite/configure +0 -14273
- data/core/vendor/src/crfsuite/configure.in +0 -149
- data/core/vendor/src/crfsuite/crfsuite.sln +0 -42
- data/core/vendor/src/crfsuite/depcomp +0 -630
- data/core/vendor/src/crfsuite/example/chunking.py +0 -49
- data/core/vendor/src/crfsuite/example/crfutils.py +0 -179
- data/core/vendor/src/crfsuite/example/ner.py +0 -270
- data/core/vendor/src/crfsuite/example/pos.py +0 -78
- data/core/vendor/src/crfsuite/example/template.py +0 -88
- data/core/vendor/src/crfsuite/frontend/Makefile.am +0 -29
- data/core/vendor/src/crfsuite/frontend/Makefile.in +0 -640
- data/core/vendor/src/crfsuite/frontend/dump.c +0 -116
- data/core/vendor/src/crfsuite/frontend/frontend.vcxproj +0 -129
- data/core/vendor/src/crfsuite/frontend/iwa.c +0 -273
- data/core/vendor/src/crfsuite/frontend/iwa.h +0 -65
- data/core/vendor/src/crfsuite/frontend/learn.c +0 -439
- data/core/vendor/src/crfsuite/frontend/main.c +0 -137
- data/core/vendor/src/crfsuite/frontend/option.c +0 -93
- data/core/vendor/src/crfsuite/frontend/option.h +0 -86
- data/core/vendor/src/crfsuite/frontend/readdata.h +0 -38
- data/core/vendor/src/crfsuite/frontend/reader.c +0 -136
- data/core/vendor/src/crfsuite/frontend/tag.c +0 -427
- data/core/vendor/src/crfsuite/genbinary.sh.in +0 -15
- data/core/vendor/src/crfsuite/include/Makefile.am +0 -11
- data/core/vendor/src/crfsuite/include/Makefile.in +0 -461
- data/core/vendor/src/crfsuite/include/crfsuite.h +0 -1063
- data/core/vendor/src/crfsuite/include/crfsuite.hpp +0 -555
- data/core/vendor/src/crfsuite/include/crfsuite_api.hpp +0 -400
- data/core/vendor/src/crfsuite/include/os.h +0 -61
- data/core/vendor/src/crfsuite/install-sh +0 -520
- data/core/vendor/src/crfsuite/lib/cqdb/COPYING +0 -28
- data/core/vendor/src/crfsuite/lib/cqdb/Makefile.am +0 -21
- data/core/vendor/src/crfsuite/lib/cqdb/Makefile.in +0 -549
- data/core/vendor/src/crfsuite/lib/cqdb/cqdb.vcxproj +0 -86
- data/core/vendor/src/crfsuite/lib/cqdb/include/cqdb.h +0 -524
- data/core/vendor/src/crfsuite/lib/cqdb/src/cqdb.c +0 -587
- data/core/vendor/src/crfsuite/lib/cqdb/src/lookup3.c +0 -976
- data/core/vendor/src/crfsuite/lib/crf/Makefile.am +0 -46
- data/core/vendor/src/crfsuite/lib/crf/Makefile.in +0 -721
- data/core/vendor/src/crfsuite/lib/crf/crf.vcxproj +0 -216
- data/core/vendor/src/crfsuite/lib/crf/src/crf1d.h +0 -353
- data/core/vendor/src/crfsuite/lib/crf/src/crf1d_context.c +0 -705
- data/core/vendor/src/crfsuite/lib/crf/src/crf1d_encode.c +0 -943
- data/core/vendor/src/crfsuite/lib/crf/src/crf1d_feature.c +0 -352
- data/core/vendor/src/crfsuite/lib/crf/src/crf1d_model.c +0 -994
- data/core/vendor/src/crfsuite/lib/crf/src/crf1d_tag.c +0 -550
- data/core/vendor/src/crfsuite/lib/crf/src/crfsuite.c +0 -492
- data/core/vendor/src/crfsuite/lib/crf/src/crfsuite_internal.h +0 -236
- data/core/vendor/src/crfsuite/lib/crf/src/crfsuite_train.c +0 -272
- data/core/vendor/src/crfsuite/lib/crf/src/dataset.c +0 -106
- data/core/vendor/src/crfsuite/lib/crf/src/dictionary.c +0 -118
- data/core/vendor/src/crfsuite/lib/crf/src/holdout.c +0 -80
- data/core/vendor/src/crfsuite/lib/crf/src/logging.c +0 -91
- data/core/vendor/src/crfsuite/lib/crf/src/logging.h +0 -48
- data/core/vendor/src/crfsuite/lib/crf/src/params.c +0 -335
- data/core/vendor/src/crfsuite/lib/crf/src/params.h +0 -80
- data/core/vendor/src/crfsuite/lib/crf/src/quark.c +0 -172
- data/core/vendor/src/crfsuite/lib/crf/src/quark.h +0 -46
- data/core/vendor/src/crfsuite/lib/crf/src/rumavl.c +0 -1107
- data/core/vendor/src/crfsuite/lib/crf/src/rumavl.h +0 -160
- data/core/vendor/src/crfsuite/lib/crf/src/train_arow.c +0 -408
- data/core/vendor/src/crfsuite/lib/crf/src/train_averaged_perceptron.c +0 -242
- data/core/vendor/src/crfsuite/lib/crf/src/train_l2sgd.c +0 -507
- data/core/vendor/src/crfsuite/lib/crf/src/train_lbfgs.c +0 -338
- data/core/vendor/src/crfsuite/lib/crf/src/train_passive_aggressive.c +0 -435
- data/core/vendor/src/crfsuite/lib/crf/src/vecmath.h +0 -341
- data/core/vendor/src/crfsuite/ltmain.sh +0 -8413
- data/core/vendor/src/crfsuite/missing +0 -376
- data/core/vendor/src/crfsuite/swig/Makefile.am +0 -13
- data/core/vendor/src/crfsuite/swig/Makefile.in +0 -365
- data/core/vendor/src/crfsuite/swig/crfsuite.cpp +0 -2
- data/core/vendor/src/crfsuite/swig/export.i +0 -32
- data/core/vendor/src/crfsuite/swig/python/README +0 -92
- data/core/vendor/src/crfsuite/swig/python/crfsuite.py +0 -329
- data/core/vendor/src/crfsuite/swig/python/export_wrap.cpp +0 -14355
- data/core/vendor/src/crfsuite/swig/python/export_wrap.h +0 -63
- data/core/vendor/src/crfsuite/swig/python/prepare.sh +0 -9
- data/core/vendor/src/crfsuite/swig/python/sample_tag.py +0 -52
- data/core/vendor/src/crfsuite/swig/python/sample_train.py +0 -68
- data/core/vendor/src/crfsuite/swig/python/setup.py +0 -44
- data/core/vendor/src/crfsuite/win32/stdint.h +0 -679
- data/core/vendor/src/liblbfgs/AUTHORS +0 -1
- data/core/vendor/src/liblbfgs/COPYING +0 -22
- data/core/vendor/src/liblbfgs/ChangeLog +0 -120
- data/core/vendor/src/liblbfgs/INSTALL +0 -231
- data/core/vendor/src/liblbfgs/Makefile.am +0 -10
- data/core/vendor/src/liblbfgs/Makefile.in +0 -638
- data/core/vendor/src/liblbfgs/NEWS +0 -0
- data/core/vendor/src/liblbfgs/README +0 -71
- data/core/vendor/src/liblbfgs/aclocal.m4 +0 -6985
- data/core/vendor/src/liblbfgs/autogen.sh +0 -38
- data/core/vendor/src/liblbfgs/config.guess +0 -1411
- data/core/vendor/src/liblbfgs/config.h.in +0 -64
- data/core/vendor/src/liblbfgs/config.sub +0 -1500
- data/core/vendor/src/liblbfgs/configure +0 -21146
- data/core/vendor/src/liblbfgs/configure.in +0 -107
- data/core/vendor/src/liblbfgs/depcomp +0 -522
- data/core/vendor/src/liblbfgs/include/lbfgs.h +0 -745
- data/core/vendor/src/liblbfgs/install-sh +0 -322
- data/core/vendor/src/liblbfgs/lbfgs.sln +0 -26
- data/core/vendor/src/liblbfgs/lib/Makefile.am +0 -24
- data/core/vendor/src/liblbfgs/lib/Makefile.in +0 -499
- data/core/vendor/src/liblbfgs/lib/arithmetic_ansi.h +0 -133
- data/core/vendor/src/liblbfgs/lib/arithmetic_sse_double.h +0 -294
- data/core/vendor/src/liblbfgs/lib/arithmetic_sse_float.h +0 -298
- data/core/vendor/src/liblbfgs/lib/lbfgs.c +0 -1371
- data/core/vendor/src/liblbfgs/lib/lib.vcxproj +0 -95
- data/core/vendor/src/liblbfgs/ltmain.sh +0 -6426
- data/core/vendor/src/liblbfgs/missing +0 -353
- data/core/vendor/src/liblbfgs/sample/Makefile.am +0 -15
- data/core/vendor/src/liblbfgs/sample/Makefile.in +0 -433
- data/core/vendor/src/liblbfgs/sample/sample.c +0 -81
- data/core/vendor/src/liblbfgs/sample/sample.cpp +0 -126
- data/core/vendor/src/liblbfgs/sample/sample.vcxproj +0 -105
- data/core/vendor/src/svm_light/LICENSE.txt +0 -59
- data/core/vendor/src/svm_light/Makefile +0 -105
- data/core/vendor/src/svm_light/kernel.h +0 -40
- data/core/vendor/src/svm_light/svm_classify.c +0 -197
- data/core/vendor/src/svm_light/svm_common.c +0 -985
- data/core/vendor/src/svm_light/svm_common.h +0 -301
- data/core/vendor/src/svm_light/svm_hideo.c +0 -1062
- data/core/vendor/src/svm_light/svm_learn.c +0 -4147
- data/core/vendor/src/svm_light/svm_learn.h +0 -169
- data/core/vendor/src/svm_light/svm_learn_main.c +0 -397
- data/core/vendor/src/svm_light/svm_loqo.c +0 -211
- data/task/c.rake +0 -36
- data/task/submodules.rake +0 -5
@@ -1,705 +0,0 @@
|
|
1
|
-
/*
|
2
|
-
* CRF1d context (forward-backward, viterbi, etc).
|
3
|
-
*
|
4
|
-
* Copyright (c) 2007-2010, Naoaki Okazaki
|
5
|
-
* All rights reserved.
|
6
|
-
*
|
7
|
-
* Redistribution and use in source and binary forms, with or without
|
8
|
-
* modification, are permitted provided that the following conditions are met:
|
9
|
-
* * Redistributions of source code must retain the above copyright
|
10
|
-
* notice, this list of conditions and the following disclaimer.
|
11
|
-
* * Redistributions in binary form must reproduce the above copyright
|
12
|
-
* notice, this list of conditions and the following disclaimer in the
|
13
|
-
* documentation and/or other materials provided with the distribution.
|
14
|
-
* * Neither the names of the authors nor the names of its contributors
|
15
|
-
* may be used to endorse or promote products derived from this
|
16
|
-
* software without specific prior written permission.
|
17
|
-
*
|
18
|
-
* THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
|
19
|
-
* "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
|
20
|
-
* LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
21
|
-
* A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
|
22
|
-
* OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
|
23
|
-
* EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
|
24
|
-
* PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
|
25
|
-
* PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
|
26
|
-
* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
|
27
|
-
* NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
|
28
|
-
* SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
|
29
|
-
*/
|
30
|
-
|
31
|
-
/* $Id$ */
|
32
|
-
|
33
|
-
#ifdef HAVE_CONFIG_H
|
34
|
-
#include <config.h>
|
35
|
-
#endif/*HAVE_CONFIG_H*/
|
36
|
-
|
37
|
-
#include <os.h>
|
38
|
-
|
39
|
-
#include <float.h>
|
40
|
-
#include <math.h>
|
41
|
-
#include <stdio.h>
|
42
|
-
#include <stdlib.h>
|
43
|
-
|
44
|
-
#include <crfsuite.h>
|
45
|
-
|
46
|
-
#include "crf1d.h"
|
47
|
-
#include "vecmath.h"
|
48
|
-
|
49
|
-
|
50
|
-
|
51
|
-
crf1d_context_t* crf1dc_new(int flag, int L, int T)
|
52
|
-
{
|
53
|
-
int ret = 0;
|
54
|
-
crf1d_context_t* ctx = NULL;
|
55
|
-
|
56
|
-
ctx = (crf1d_context_t*)calloc(1, sizeof(crf1d_context_t));
|
57
|
-
if (ctx != NULL) {
|
58
|
-
ctx->flag = flag;
|
59
|
-
ctx->num_labels = L;
|
60
|
-
|
61
|
-
ctx->trans = (floatval_t*)calloc(L * L, sizeof(floatval_t));
|
62
|
-
if (ctx->trans == NULL) goto error_exit;
|
63
|
-
|
64
|
-
if (ctx->flag & CTXF_MARGINALS) {
|
65
|
-
ctx->exp_trans = (floatval_t*)_aligned_malloc((L * L + 4) * sizeof(floatval_t), 16);
|
66
|
-
if (ctx->exp_trans == NULL) goto error_exit;
|
67
|
-
ctx->mexp_trans = (floatval_t*)calloc(L * L, sizeof(floatval_t));
|
68
|
-
if (ctx->mexp_trans == NULL) goto error_exit;
|
69
|
-
}
|
70
|
-
|
71
|
-
if (ret = crf1dc_set_num_items(ctx, T)) {
|
72
|
-
goto error_exit;
|
73
|
-
}
|
74
|
-
|
75
|
-
/* T gives the 'hint' for maximum length of items. */
|
76
|
-
ctx->num_items = 0;
|
77
|
-
}
|
78
|
-
|
79
|
-
return ctx;
|
80
|
-
|
81
|
-
error_exit:
|
82
|
-
crf1dc_delete(ctx);
|
83
|
-
return NULL;
|
84
|
-
}
|
85
|
-
|
86
|
-
int crf1dc_set_num_items(crf1d_context_t* ctx, int T)
|
87
|
-
{
|
88
|
-
const int L = ctx->num_labels;
|
89
|
-
|
90
|
-
ctx->num_items = T;
|
91
|
-
|
92
|
-
if (ctx->cap_items < T) {
|
93
|
-
free(ctx->backward_edge);
|
94
|
-
free(ctx->mexp_state);
|
95
|
-
_aligned_free(ctx->exp_state);
|
96
|
-
free(ctx->scale_factor);
|
97
|
-
free(ctx->row);
|
98
|
-
free(ctx->beta_score);
|
99
|
-
free(ctx->alpha_score);
|
100
|
-
|
101
|
-
ctx->alpha_score = (floatval_t*)calloc(T * L, sizeof(floatval_t));
|
102
|
-
if (ctx->alpha_score == NULL) return CRFSUITEERR_OUTOFMEMORY;
|
103
|
-
ctx->beta_score = (floatval_t*)calloc(T * L, sizeof(floatval_t));
|
104
|
-
if (ctx->beta_score == NULL) return CRFSUITEERR_OUTOFMEMORY;
|
105
|
-
ctx->scale_factor = (floatval_t*)calloc(T, sizeof(floatval_t));
|
106
|
-
if (ctx->scale_factor == NULL) return CRFSUITEERR_OUTOFMEMORY;
|
107
|
-
ctx->row = (floatval_t*)calloc(L, sizeof(floatval_t));
|
108
|
-
if (ctx->row == NULL) return CRFSUITEERR_OUTOFMEMORY;
|
109
|
-
|
110
|
-
if (ctx->flag & CTXF_VITERBI) {
|
111
|
-
ctx->backward_edge = (int*)calloc(T * L, sizeof(int));
|
112
|
-
if (ctx->backward_edge == NULL) return CRFSUITEERR_OUTOFMEMORY;
|
113
|
-
}
|
114
|
-
|
115
|
-
ctx->state = (floatval_t*)calloc(T * L, sizeof(floatval_t));
|
116
|
-
if (ctx->state == NULL) return CRFSUITEERR_OUTOFMEMORY;
|
117
|
-
|
118
|
-
if (ctx->flag & CTXF_MARGINALS) {
|
119
|
-
ctx->exp_state = (floatval_t*)_aligned_malloc((T * L + 4) * sizeof(floatval_t), 16);
|
120
|
-
if (ctx->exp_state == NULL) return CRFSUITEERR_OUTOFMEMORY;
|
121
|
-
ctx->mexp_state = (floatval_t*)calloc(T * L, sizeof(floatval_t));
|
122
|
-
if (ctx->mexp_state == NULL) return CRFSUITEERR_OUTOFMEMORY;
|
123
|
-
}
|
124
|
-
|
125
|
-
ctx->cap_items = T;
|
126
|
-
}
|
127
|
-
|
128
|
-
return 0;
|
129
|
-
}
|
130
|
-
|
131
|
-
void crf1dc_delete(crf1d_context_t* ctx)
|
132
|
-
{
|
133
|
-
if (ctx != NULL) {
|
134
|
-
free(ctx->backward_edge);
|
135
|
-
free(ctx->mexp_state);
|
136
|
-
_aligned_free(ctx->exp_state);
|
137
|
-
free(ctx->state);
|
138
|
-
free(ctx->scale_factor);
|
139
|
-
free(ctx->row);
|
140
|
-
free(ctx->beta_score);
|
141
|
-
free(ctx->alpha_score);
|
142
|
-
free(ctx->mexp_trans);
|
143
|
-
_aligned_free(ctx->exp_trans);
|
144
|
-
free(ctx->trans);
|
145
|
-
}
|
146
|
-
free(ctx);
|
147
|
-
}
|
148
|
-
|
149
|
-
void crf1dc_reset(crf1d_context_t* ctx, int flag)
|
150
|
-
{
|
151
|
-
const int T = ctx->num_items;
|
152
|
-
const int L = ctx->num_labels;
|
153
|
-
|
154
|
-
if (flag & RF_STATE) {
|
155
|
-
veczero(ctx->state, T*L);
|
156
|
-
}
|
157
|
-
if (flag & RF_TRANS) {
|
158
|
-
veczero(ctx->trans, L*L);
|
159
|
-
}
|
160
|
-
|
161
|
-
if (ctx->flag & CTXF_MARGINALS) {
|
162
|
-
veczero(ctx->mexp_state, T*L);
|
163
|
-
veczero(ctx->mexp_trans, L*L);
|
164
|
-
ctx->log_norm = 0;
|
165
|
-
}
|
166
|
-
}
|
167
|
-
|
168
|
-
void crf1dc_exp_state(crf1d_context_t* ctx)
|
169
|
-
{
|
170
|
-
const int T = ctx->num_items;
|
171
|
-
const int L = ctx->num_labels;
|
172
|
-
|
173
|
-
veccopy(ctx->exp_state, ctx->state, L * T);
|
174
|
-
vecexp(ctx->exp_state, L * T);
|
175
|
-
}
|
176
|
-
|
177
|
-
void crf1dc_exp_transition(crf1d_context_t* ctx)
|
178
|
-
{
|
179
|
-
const int L = ctx->num_labels;
|
180
|
-
|
181
|
-
veccopy(ctx->exp_trans, ctx->trans, L * L);
|
182
|
-
vecexp(ctx->exp_trans, L * L);
|
183
|
-
}
|
184
|
-
|
185
|
-
void crf1dc_alpha_score(crf1d_context_t* ctx)
|
186
|
-
{
|
187
|
-
int i, t;
|
188
|
-
floatval_t sum, *cur = NULL;
|
189
|
-
floatval_t *scale = &ctx->scale_factor[0];
|
190
|
-
const floatval_t *prev = NULL, *trans = NULL, *state = NULL;
|
191
|
-
const int T = ctx->num_items;
|
192
|
-
const int L = ctx->num_labels;
|
193
|
-
|
194
|
-
/* Compute the alpha scores on nodes (0, *).
|
195
|
-
alpha[0][j] = state[0][j]
|
196
|
-
*/
|
197
|
-
cur = ALPHA_SCORE(ctx, 0);
|
198
|
-
state = EXP_STATE_SCORE(ctx, 0);
|
199
|
-
veccopy(cur, state, L);
|
200
|
-
sum = vecsum(cur, L);
|
201
|
-
*scale = (sum != 0.) ? 1. / sum : 1.;
|
202
|
-
vecscale(cur, *scale, L);
|
203
|
-
++scale;
|
204
|
-
|
205
|
-
/* Compute the alpha scores on nodes (t, *).
|
206
|
-
alpha[t][j] = state[t][j] * \sum_{i} alpha[t-1][i] * trans[i][j]
|
207
|
-
*/
|
208
|
-
for (t = 1;t < T;++t) {
|
209
|
-
prev = ALPHA_SCORE(ctx, t-1);
|
210
|
-
cur = ALPHA_SCORE(ctx, t);
|
211
|
-
state = EXP_STATE_SCORE(ctx, t);
|
212
|
-
|
213
|
-
veczero(cur, L);
|
214
|
-
for (i = 0;i < L;++i) {
|
215
|
-
trans = EXP_TRANS_SCORE(ctx, i);
|
216
|
-
vecaadd(cur, prev[i], trans, L);
|
217
|
-
}
|
218
|
-
vecmul(cur, state, L);
|
219
|
-
sum = vecsum(cur, L);
|
220
|
-
*scale = (sum != 0.) ? 1. / sum : 1.;
|
221
|
-
vecscale(cur, *scale, L);
|
222
|
-
++scale;
|
223
|
-
}
|
224
|
-
|
225
|
-
/* Compute the logarithm of the normalization factor here.
|
226
|
-
norm = 1. / (C[0] * C[1] ... * C[T-1])
|
227
|
-
log(norm) = - \sum_{t = 0}^{T-1} log(C[t]).
|
228
|
-
*/
|
229
|
-
ctx->log_norm = -vecsumlog(ctx->scale_factor, T);
|
230
|
-
}
|
231
|
-
|
232
|
-
void crf1dc_beta_score(crf1d_context_t* ctx)
|
233
|
-
{
|
234
|
-
int i, t;
|
235
|
-
floatval_t *cur = NULL;
|
236
|
-
floatval_t *row = ctx->row;
|
237
|
-
const floatval_t *next = NULL, *state = NULL, *trans = NULL;
|
238
|
-
const int T = ctx->num_items;
|
239
|
-
const int L = ctx->num_labels;
|
240
|
-
const floatval_t *scale = &ctx->scale_factor[T-1];
|
241
|
-
|
242
|
-
/* Compute the beta scores at (T-1, *). */
|
243
|
-
cur = BETA_SCORE(ctx, T-1);
|
244
|
-
vecset(cur, *scale, L);
|
245
|
-
--scale;
|
246
|
-
|
247
|
-
/* Compute the beta scores at (t, *). */
|
248
|
-
for (t = T-2;0 <= t;--t) {
|
249
|
-
cur = BETA_SCORE(ctx, t);
|
250
|
-
next = BETA_SCORE(ctx, t+1);
|
251
|
-
state = EXP_STATE_SCORE(ctx, t+1);
|
252
|
-
|
253
|
-
veccopy(row, next, L);
|
254
|
-
vecmul(row, state, L);
|
255
|
-
|
256
|
-
/* Compute the beta score at (t, i). */
|
257
|
-
for (i = 0;i < L;++i) {
|
258
|
-
trans = EXP_TRANS_SCORE(ctx, i);
|
259
|
-
cur[i] = vecdot(trans, row, L);
|
260
|
-
}
|
261
|
-
vecscale(cur, *scale, L);
|
262
|
-
--scale;
|
263
|
-
}
|
264
|
-
}
|
265
|
-
|
266
|
-
void crf1dc_marginals(crf1d_context_t* ctx)
|
267
|
-
{
|
268
|
-
int i, j, t;
|
269
|
-
const int T = ctx->num_items;
|
270
|
-
const int L = ctx->num_labels;
|
271
|
-
|
272
|
-
/*
|
273
|
-
Compute the model expectations of states.
|
274
|
-
p(t,i) = fwd[t][i] * bwd[t][i] / norm
|
275
|
-
= (1. / C[t]) * fwd'[t][i] * bwd'[t][i]
|
276
|
-
*/
|
277
|
-
for (t = 0;t < T;++t) {
|
278
|
-
floatval_t *fwd = ALPHA_SCORE(ctx, t);
|
279
|
-
floatval_t *bwd = BETA_SCORE(ctx, t);
|
280
|
-
floatval_t *prob = STATE_MEXP(ctx, t);
|
281
|
-
veccopy(prob, fwd, L);
|
282
|
-
vecmul(prob, bwd, L);
|
283
|
-
vecscale(prob, 1. / ctx->scale_factor[t], L);
|
284
|
-
}
|
285
|
-
|
286
|
-
/*
|
287
|
-
Compute the model expectations of transitions.
|
288
|
-
p(t,i,t+1,j)
|
289
|
-
= fwd[t][i] * edge[i][j] * state[t+1][j] * bwd[t+1][j] / norm
|
290
|
-
= (fwd'[t][i] / (C[0] ... C[t])) * edge[i][j] * state[t+1][j] * (bwd'[t+1][j] / (C[t+1] ... C[T-1])) * (C[0] * ... * C[T-1])
|
291
|
-
= fwd'[t][i] * edge[i][j] * state[t+1][j] * bwd'[t+1][j]
|
292
|
-
The model expectation of a transition (i -> j) is the sum of the marginal
|
293
|
-
probabilities p(t,i,t+1,j) over t.
|
294
|
-
*/
|
295
|
-
for (t = 0;t < T-1;++t) {
|
296
|
-
floatval_t *fwd = ALPHA_SCORE(ctx, t);
|
297
|
-
floatval_t *state = EXP_STATE_SCORE(ctx, t+1);
|
298
|
-
floatval_t *bwd = BETA_SCORE(ctx, t+1);
|
299
|
-
floatval_t *row = ctx->row;
|
300
|
-
|
301
|
-
/* row[j] = state[t+1][j] * bwd'[t+1][j] */
|
302
|
-
veccopy(row, bwd, L);
|
303
|
-
vecmul(row, state, L);
|
304
|
-
|
305
|
-
for (i = 0;i < L;++i) {
|
306
|
-
floatval_t *edge = EXP_TRANS_SCORE(ctx, i);
|
307
|
-
floatval_t *prob = TRANS_MEXP(ctx, i);
|
308
|
-
for (j = 0;j < L;++j) {
|
309
|
-
prob[j] += fwd[i] * edge[j] * row[j];
|
310
|
-
}
|
311
|
-
}
|
312
|
-
}
|
313
|
-
}
|
314
|
-
|
315
|
-
floatval_t crf1dc_marginal_point(crf1d_context_t *ctx, int l, int t)
|
316
|
-
{
|
317
|
-
floatval_t *fwd = ALPHA_SCORE(ctx, t);
|
318
|
-
floatval_t *bwd = BETA_SCORE(ctx, t);
|
319
|
-
return fwd[l] * bwd[l] / ctx->scale_factor[t];
|
320
|
-
}
|
321
|
-
|
322
|
-
floatval_t crf1dc_marginal_path(crf1d_context_t *ctx, const int *path, int begin, int end)
|
323
|
-
{
|
324
|
-
int t;
|
325
|
-
/*
|
326
|
-
Compute the marginal probability of a (partial) path.
|
327
|
-
a = path[begin], b = path[begin+1], ..., y = path[end-2], z = path[end-1]
|
328
|
-
fwd[begin][a] = (fwd'[begin][a] / (C[0] ... C[begin])
|
329
|
-
bwd[end-1][z] = (bwd'[end-1][z] / (C[end-1] ... C[T-1]))
|
330
|
-
norm = 1 / (C[0] * ... * C[T-1])
|
331
|
-
p(a, b, ..., z)
|
332
|
-
= fwd[begin][a] * edge[a][b] * state[begin+1][b] * ... * edge[y][z] * state[end-1][z] * bwd[end-1][z] / norm
|
333
|
-
= fwd'[begin][a] * edge[a][b] * state[begin+1][b] * ... * edge[y][z] * state[end-1][z] * bwd'[end-1][z] * (C[begin+1] * ... * C[end-2])
|
334
|
-
*/
|
335
|
-
floatval_t *fwd = ALPHA_SCORE(ctx, begin);
|
336
|
-
floatval_t *bwd = BETA_SCORE(ctx, end-1);
|
337
|
-
floatval_t prob = fwd[path[begin]] * bwd[path[end-1]] / ctx->scale_factor[begin];
|
338
|
-
|
339
|
-
for (t = begin;t < end-1;++t) {
|
340
|
-
floatval_t *state = EXP_STATE_SCORE(ctx, t+1);
|
341
|
-
floatval_t *edge = EXP_TRANS_SCORE(ctx, path[t]);
|
342
|
-
prob *= (edge[path[t+1]] * state[path[t+1]] * ctx->scale_factor[t]);
|
343
|
-
}
|
344
|
-
|
345
|
-
return prob;
|
346
|
-
}
|
347
|
-
|
348
|
-
#if 0
|
349
|
-
/* Sigh, this was found to be slower than the forward-backward algorithm. */
|
350
|
-
|
351
|
-
#define ADJACENCY(ctx, i) \
|
352
|
-
(&MATRIX(ctx->adj, ctx->num_labels, 0, i))
|
353
|
-
|
354
|
-
void crf1dc_marginal_without_beta(crf1d_context_t* ctx)
|
355
|
-
{
|
356
|
-
int i, j, t;
|
357
|
-
floatval_t *prob = NULL;
|
358
|
-
floatval_t *row = ctx->row;
|
359
|
-
const floatval_t *fwd = NULL;
|
360
|
-
const int T = ctx->num_items;
|
361
|
-
const int L = ctx->num_labels;
|
362
|
-
|
363
|
-
/*
|
364
|
-
Compute marginal probabilities of states at T-1
|
365
|
-
p(T-1,j) = fwd'[T-1][j]
|
366
|
-
*/
|
367
|
-
fwd = ALPHA_SCORE(ctx, T-1);
|
368
|
-
prob = STATE_MEXP(ctx, T-1);
|
369
|
-
veccopy(prob, fwd, L);
|
370
|
-
|
371
|
-
/*
|
372
|
-
Repeat the following computation for t = T-1,T-2, ..., 1.
|
373
|
-
1) Compute p(t-1,i,t,j) using p(t,j)
|
374
|
-
2) Compute p(t,i) using p(t-1,i,t,j)
|
375
|
-
*/
|
376
|
-
for (t = T-1;0 < t;--t) {
|
377
|
-
fwd = ALPHA_SCORE(ctx, t-1);
|
378
|
-
prob = STATE_MEXP(ctx, t);
|
379
|
-
|
380
|
-
veczero(ctx->adj, L*L);
|
381
|
-
veczero(row, L);
|
382
|
-
|
383
|
-
/*
|
384
|
-
Compute adj[i][j] and row[j].
|
385
|
-
adj[i][j] = fwd'[t-1][i] * edge[i][j]
|
386
|
-
row[j] = \sum_{i} adj[i][j]
|
387
|
-
*/
|
388
|
-
for (i = 0;i < L;++i) {
|
389
|
-
floatval_t *adj = ADJACENCY(ctx, i);
|
390
|
-
floatval_t *edge = EXP_TRANS_SCORE(ctx, i);
|
391
|
-
vecaadd(adj, fwd[i], edge, L);
|
392
|
-
vecadd(row, adj, L);
|
393
|
-
}
|
394
|
-
|
395
|
-
/*
|
396
|
-
Find z such that z * \sum_{i] adj[i][j] = p(t,j).
|
397
|
-
Thus, z = p(t,j) / row[j]; we overwrite row with z.
|
398
|
-
*/
|
399
|
-
vecinv(row, L);
|
400
|
-
vecmul(row, prob, L);
|
401
|
-
|
402
|
-
/*
|
403
|
-
Apply the partition factor z (row[j]) to adj[i][j].
|
404
|
-
*/
|
405
|
-
for (i = 0;i < L;++i) {
|
406
|
-
floatval_t *adj = ADJACENCY(ctx, i);
|
407
|
-
vecmul(adj, row, L);
|
408
|
-
}
|
409
|
-
|
410
|
-
/*
|
411
|
-
Now that adj[i][j] presents p(t-1,i,t,j),
|
412
|
-
accumulate model expectations of transitions.
|
413
|
-
*/
|
414
|
-
for (i = 0;i < L;++i) {
|
415
|
-
floatval_t *adj = ADJACENCY(ctx, i);
|
416
|
-
floatval_t *prob = TRANS_MEXP(ctx, i);
|
417
|
-
vecadd(prob, adj, L);
|
418
|
-
}
|
419
|
-
|
420
|
-
/*
|
421
|
-
Compute the marginal probability of states at t-1.
|
422
|
-
p(t-1,i) = \sum_{j} p(t-1,i,t,j)
|
423
|
-
*/
|
424
|
-
prob = STATE_MEXP(ctx, t-1);
|
425
|
-
for (i = 0;i < L;++i) {
|
426
|
-
floatval_t *adj = ADJACENCY(ctx, i);
|
427
|
-
prob[i] = vecsum(adj, L);
|
428
|
-
}
|
429
|
-
}
|
430
|
-
}
|
431
|
-
#endif
|
432
|
-
|
433
|
-
floatval_t crf1dc_score(crf1d_context_t* ctx, const int *labels)
|
434
|
-
{
|
435
|
-
int i, j, t;
|
436
|
-
floatval_t ret = 0;
|
437
|
-
const floatval_t *state = NULL, *cur = NULL, *trans = NULL;
|
438
|
-
const int T = ctx->num_items;
|
439
|
-
const int L = ctx->num_labels;
|
440
|
-
|
441
|
-
/* Stay at (0, labels[0]). */
|
442
|
-
i = labels[0];
|
443
|
-
state = STATE_SCORE(ctx, 0);
|
444
|
-
ret = state[i];
|
445
|
-
|
446
|
-
/* Loop over the rest of items. */
|
447
|
-
for (t = 1;t < T;++t) {
|
448
|
-
j = labels[t];
|
449
|
-
trans = TRANS_SCORE(ctx, i);
|
450
|
-
state = STATE_SCORE(ctx, t);
|
451
|
-
|
452
|
-
/* Transit from (t-1, i) to (t, j). */
|
453
|
-
ret += trans[j];
|
454
|
-
ret += state[j];
|
455
|
-
i = j;
|
456
|
-
}
|
457
|
-
return ret;
|
458
|
-
}
|
459
|
-
|
460
|
-
floatval_t crf1dc_lognorm(crf1d_context_t* ctx)
|
461
|
-
{
|
462
|
-
return ctx->log_norm;
|
463
|
-
}
|
464
|
-
|
465
|
-
floatval_t crf1dc_viterbi(crf1d_context_t* ctx, int *labels)
|
466
|
-
{
|
467
|
-
int i, j, t;
|
468
|
-
int *back = NULL;
|
469
|
-
floatval_t max_score, score, *cur = NULL;
|
470
|
-
const floatval_t *prev = NULL, *state = NULL, *trans = NULL;
|
471
|
-
const int T = ctx->num_items;
|
472
|
-
const int L = ctx->num_labels;
|
473
|
-
|
474
|
-
/*
|
475
|
-
This function assumes state and trans scores to be in the logarithm domain.
|
476
|
-
*/
|
477
|
-
|
478
|
-
/* Compute the scores at (0, *). */
|
479
|
-
cur = ALPHA_SCORE(ctx, 0);
|
480
|
-
state = STATE_SCORE(ctx, 0);
|
481
|
-
for (j = 0;j < L;++j) {
|
482
|
-
cur[j] = state[j];
|
483
|
-
}
|
484
|
-
|
485
|
-
/* Compute the scores at (t, *). */
|
486
|
-
for (t = 1;t < T;++t) {
|
487
|
-
prev = ALPHA_SCORE(ctx, t-1);
|
488
|
-
cur = ALPHA_SCORE(ctx, t);
|
489
|
-
state = STATE_SCORE(ctx, t);
|
490
|
-
back = BACKWARD_EDGE_AT(ctx, t);
|
491
|
-
|
492
|
-
/* Compute the score of (t, j). */
|
493
|
-
for (j = 0;j < L;++j) {
|
494
|
-
max_score = -FLOAT_MAX;
|
495
|
-
|
496
|
-
for (i = 0;i < L;++i) {
|
497
|
-
/* Transit from (t-1, i) to (t, j). */
|
498
|
-
trans = TRANS_SCORE(ctx, i);
|
499
|
-
score = prev[i] + trans[j];
|
500
|
-
|
501
|
-
/* Store this path if it has the maximum score. */
|
502
|
-
if (max_score < score) {
|
503
|
-
max_score = score;
|
504
|
-
/* Backward link (#t, #j) -> (#t-1, #i). */
|
505
|
-
back[j] = i;
|
506
|
-
}
|
507
|
-
}
|
508
|
-
/* Add the state score on (t, j). */
|
509
|
-
cur[j] = max_score + state[j];
|
510
|
-
}
|
511
|
-
}
|
512
|
-
|
513
|
-
/* Find the node (#T, #i) that reaches EOS with the maximum score. */
|
514
|
-
max_score = -FLOAT_MAX;
|
515
|
-
prev = ALPHA_SCORE(ctx, T-1);
|
516
|
-
for (i = 0;i < L;++i) {
|
517
|
-
if (max_score < prev[i]) {
|
518
|
-
max_score = prev[i];
|
519
|
-
labels[T-1] = i; /* Tag the item #T. */
|
520
|
-
}
|
521
|
-
}
|
522
|
-
|
523
|
-
/* Tag labels by tracing the backward links. */
|
524
|
-
for (t = T-2;0 <= t;--t) {
|
525
|
-
back = BACKWARD_EDGE_AT(ctx, t+1);
|
526
|
-
labels[t] = back[labels[t+1]];
|
527
|
-
}
|
528
|
-
|
529
|
-
/* Return the maximum score (without the normalization factor subtracted). */
|
530
|
-
return max_score;
|
531
|
-
}
|
532
|
-
|
533
|
-
static void check_values(FILE *fp, floatval_t cv, floatval_t tv)
|
534
|
-
{
|
535
|
-
if (fabs(cv - tv) < 1e-9) {
|
536
|
-
fprintf(fp, "OK (%f)\n", cv);
|
537
|
-
} else {
|
538
|
-
fprintf(fp, "FAIL: %f (%f)\n", cv, tv);
|
539
|
-
}
|
540
|
-
}
|
541
|
-
|
542
|
-
void crf1dc_debug_context(FILE *fp)
|
543
|
-
{
|
544
|
-
int y1, y2, y3;
|
545
|
-
floatval_t norm = 0;
|
546
|
-
const int L = 3;
|
547
|
-
const int T = 3;
|
548
|
-
crf1d_context_t *ctx = crf1dc_new(CTXF_MARGINALS, L, T);
|
549
|
-
floatval_t *trans = NULL, *state = NULL;
|
550
|
-
floatval_t scores[3][3][3];
|
551
|
-
int labels[3];
|
552
|
-
|
553
|
-
/* Initialize the state scores. */
|
554
|
-
state = EXP_STATE_SCORE(ctx, 0);
|
555
|
-
state[0] = .4; state[1] = .5; state[2] = .1;
|
556
|
-
state = EXP_STATE_SCORE(ctx, 1);
|
557
|
-
state[0] = .4; state[1] = .1; state[2] = .5;
|
558
|
-
state = EXP_STATE_SCORE(ctx, 2);
|
559
|
-
state[0] = .4; state[1] = .1; state[2] = .5;
|
560
|
-
|
561
|
-
/* Initialize the transition scores. */
|
562
|
-
trans = EXP_TRANS_SCORE(ctx, 0);
|
563
|
-
trans[0] = .3; trans[1] = .1; trans[2] = .4;
|
564
|
-
trans = EXP_TRANS_SCORE(ctx, 1);
|
565
|
-
trans[0] = .6; trans[1] = .2; trans[2] = .1;
|
566
|
-
trans = EXP_TRANS_SCORE(ctx, 2);
|
567
|
-
trans[0] = .5; trans[1] = .2; trans[2] = .1;
|
568
|
-
|
569
|
-
ctx->num_items = ctx->cap_items;
|
570
|
-
crf1dc_alpha_score(ctx);
|
571
|
-
crf1dc_beta_score(ctx);
|
572
|
-
|
573
|
-
/* Compute the score of every label sequence. */
|
574
|
-
for (y1 = 0;y1 < L;++y1) {
|
575
|
-
floatval_t s1 = EXP_STATE_SCORE(ctx, 0)[y1];
|
576
|
-
for (y2 = 0;y2 < L;++y2) {
|
577
|
-
floatval_t s2 = s1;
|
578
|
-
s2 *= EXP_TRANS_SCORE(ctx, y1)[y2];
|
579
|
-
s2 *= EXP_STATE_SCORE(ctx, 1)[y2];
|
580
|
-
for (y3 = 0;y3 < L;++y3) {
|
581
|
-
floatval_t s3 = s2;
|
582
|
-
s3 *= EXP_TRANS_SCORE(ctx, y2)[y3];
|
583
|
-
s3 *= EXP_STATE_SCORE(ctx, 2)[y3];
|
584
|
-
scores[y1][y2][y3] = s3;
|
585
|
-
}
|
586
|
-
}
|
587
|
-
}
|
588
|
-
|
589
|
-
/* Compute the partition factor. */
|
590
|
-
norm = 0.;
|
591
|
-
for (y1 = 0;y1 < L;++y1) {
|
592
|
-
for (y2 = 0;y2 < L;++y2) {
|
593
|
-
for (y3 = 0;y3 < L;++y3) {
|
594
|
-
norm += scores[y1][y2][y3];
|
595
|
-
}
|
596
|
-
}
|
597
|
-
}
|
598
|
-
|
599
|
-
/* Check the partition factor. */
|
600
|
-
fprintf(fp, "Check for the partition factor... ");
|
601
|
-
check_values(fp, exp(ctx->log_norm), norm);
|
602
|
-
|
603
|
-
/* Compute the sequence probabilities. */
|
604
|
-
for (y1 = 0;y1 < L;++y1) {
|
605
|
-
for (y2 = 0;y2 < L;++y2) {
|
606
|
-
for (y3 = 0;y3 < L;++y3) {
|
607
|
-
floatval_t logp;
|
608
|
-
|
609
|
-
labels[0] = y1;
|
610
|
-
labels[1] = y2;
|
611
|
-
labels[2] = y3;
|
612
|
-
logp = crf1dc_score(ctx, labels) - crf1dc_lognorm(ctx);
|
613
|
-
|
614
|
-
fprintf(fp, "Check for the sequence %d-%d-%d... ", y1, y2, y3);
|
615
|
-
check_values(fp, exp(logp), scores[y1][y2][y3] / norm);
|
616
|
-
}
|
617
|
-
}
|
618
|
-
}
|
619
|
-
|
620
|
-
/* Compute the marginal probability at t=0 */
|
621
|
-
for (y1 = 0;y1 < L;++y1) {
|
622
|
-
floatval_t a, b, c, s = 0.;
|
623
|
-
for (y2 = 0;y2 < L;++y2) {
|
624
|
-
for (y3 = 0;y3 < L;++y3) {
|
625
|
-
s += scores[y1][y2][y3];
|
626
|
-
}
|
627
|
-
}
|
628
|
-
|
629
|
-
a = ALPHA_SCORE(ctx, 0)[y1];
|
630
|
-
b = BETA_SCORE(ctx, 0)[y1];
|
631
|
-
c = 1. / ctx->scale_factor[0];
|
632
|
-
|
633
|
-
fprintf(fp, "Check for the marginal probability (0,%d)... ", y1);
|
634
|
-
check_values(fp, a * b * c, s / norm);
|
635
|
-
}
|
636
|
-
|
637
|
-
/* Compute the marginal probability at t=1 */
|
638
|
-
for (y2 = 0;y2 < L;++y2) {
|
639
|
-
floatval_t a, b, c, s = 0.;
|
640
|
-
for (y1 = 0;y1 < L;++y1) {
|
641
|
-
for (y3 = 0;y3 < L;++y3) {
|
642
|
-
s += scores[y1][y2][y3];
|
643
|
-
}
|
644
|
-
}
|
645
|
-
|
646
|
-
a = ALPHA_SCORE(ctx, 1)[y2];
|
647
|
-
b = BETA_SCORE(ctx, 1)[y2];
|
648
|
-
c = 1. / ctx->scale_factor[1];
|
649
|
-
|
650
|
-
fprintf(fp, "Check for the marginal probability (1,%d)... ", y2);
|
651
|
-
check_values(fp, a * b * c, s / norm);
|
652
|
-
}
|
653
|
-
|
654
|
-
/* Compute the marginal probability at t=2 */
|
655
|
-
for (y3 = 0;y3 < L;++y3) {
|
656
|
-
floatval_t a, b, c, s = 0.;
|
657
|
-
for (y1 = 0;y1 < L;++y1) {
|
658
|
-
for (y2 = 0;y2 < L;++y2) {
|
659
|
-
s += scores[y1][y2][y3];
|
660
|
-
}
|
661
|
-
}
|
662
|
-
|
663
|
-
a = ALPHA_SCORE(ctx, 2)[y3];
|
664
|
-
b = BETA_SCORE(ctx, 2)[y3];
|
665
|
-
c = 1. / ctx->scale_factor[2];
|
666
|
-
|
667
|
-
fprintf(fp, "Check for the marginal probability (2,%d)... ", y3);
|
668
|
-
check_values(fp, a * b * c, s / norm);
|
669
|
-
}
|
670
|
-
|
671
|
-
/* Compute the marginal probabilities of transitions. */
|
672
|
-
for (y1 = 0;y1 < L;++y1) {
|
673
|
-
for (y2 = 0;y2 < L;++y2) {
|
674
|
-
floatval_t a, b, s, t, p = 0.;
|
675
|
-
for (y3 = 0;y3 < L;++y3) {
|
676
|
-
p += scores[y1][y2][y3];
|
677
|
-
}
|
678
|
-
|
679
|
-
a = ALPHA_SCORE(ctx, 0)[y1];
|
680
|
-
b = BETA_SCORE(ctx, 1)[y2];
|
681
|
-
s = EXP_STATE_SCORE(ctx, 1)[y2];
|
682
|
-
t = EXP_TRANS_SCORE(ctx, y1)[y2];
|
683
|
-
|
684
|
-
fprintf(fp, "Check for the marginal probability (0,%d)-(1,%d)... ", y1, y2);
|
685
|
-
check_values(fp, a * t * s * b, p / norm);
|
686
|
-
}
|
687
|
-
}
|
688
|
-
|
689
|
-
for (y2 = 0;y2 < L;++y2) {
|
690
|
-
for (y3 = 0;y3 < L;++y3) {
|
691
|
-
floatval_t a, b, s, t, p = 0.;
|
692
|
-
for (y1 = 0;y1 < L;++y1) {
|
693
|
-
p += scores[y1][y2][y3];
|
694
|
-
}
|
695
|
-
|
696
|
-
a = ALPHA_SCORE(ctx, 1)[y2];
|
697
|
-
b = BETA_SCORE(ctx, 2)[y3];
|
698
|
-
s = EXP_STATE_SCORE(ctx, 2)[y3];
|
699
|
-
t = EXP_TRANS_SCORE(ctx, y2)[y3];
|
700
|
-
|
701
|
-
fprintf(fp, "Check for the marginal probability (1,%d)-(2,%d)... ", y2, y3);
|
702
|
-
check_values(fp, a * t * s * b, p / norm);
|
703
|
-
}
|
704
|
-
}
|
705
|
-
}
|