forked from pprakhar30/MOQA
-
Notifications
You must be signed in to change notification settings - Fork 0
/
model.py
666 lines (498 loc) · 21.8 KB
/
model.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
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
import json
import numpy as np
import tensorflow as tf
from math import ceil,log
from random import randint
from collections import defaultdict, OrderedDict
class Model:
def __init__(self, k, numIter, Lambda ,corpus):
'''
Parameters
-------------
k: the rank of the matrices which approximate the matrix M in qMd'. Such that M = AB', where each A and B have rank 'k'
numIter: Number of iterations you want to train your model
Lambda: regularization parameter
corpus: an object of Corpus class
Model Parameters
-----------------
theta : parameter corresponding to pairwise similarity
RelvPar: parameter corresponding to term to term Relevance of question and review
A : parameter corresponding to the bilinear term, projects question to 'k' dimensional space
B, Y : parameter corresponding to the bilinear term, projects review to 'k' dimensional space
PredPar: parameter corresponding to term to term Relevance of answer and review
X : parameter corresponding to the bilinear term, projects answer to 'k' dimensional space
Other Attributes:
-----------------
PairwiseDim: takes bm25, bm25+ and LCS as three off the plate similarity measures and hence 3
V: voublary size of the corpus we have read
Na: Each answer matrix contains 11 answer 0th being the right answer and other 10 being the wrong answer for training purpose. Total answers hence is 11 by default.
Items: Our training is done item wise each item define its own set of documents
valid: the training validation split
valid_test: the test training split
trainQPerItem: Stores the Ids of the question used for training
validTestQ: Stores the Ids of the question in validation and test set
validTestNa: Stores the ids of the non answer to be used to evalueate the validation and test set performance during training.
validTestM: Stores the Sparse Matrix representing all the questions, answer, review, term to term similarity and pairwise similarity for each question in validation and test set
Pairwise: Stores the pairwise feature for each item.
Question: Stores the Sparse Matrix representing all the questions in a item for each item
Answer: Stores the Sparse Matrix representing all the answers (also the 10 non-answer) corresponding to a question in a item for each item
Review: Stores the Sparse Matrix representing all the reviews in a item for each item
TermtoTermR: Stores the Sparse Matrix representing all the term to term similarity between question and reviews in a item for each item
TermtoTermP: Stores the Sparse Matrix representing all the term to term similarity between answer and reviews in a item for each item
Question_I: Sparse Bit mask for Question Matrix above
Answer_I: Sparse Bit mask for Answer Matrix above
Review_I: Sparse Bit mask for Review Matrix above
pairwise, question, answer, review, question_I, answer_I, review_I, termTotermR, termTotermP: Placeholder for above matrices
'''
self.PairwiseDim = 3
self.rankDim = k
self.V = corpus.Map.V
self.Na = 11
self.Items = len(corpus.Map.ItemIDMap)
self.numIter = numIter
self.Lambda = Lambda
self.corpus = corpus
self.valid = 0.8
self.valid_test = 0.5
self.theta = None
self.RelvPar = None
self.A = None
self.B = None
self.PredPar = None
self.X = None
self.Y = None
self.loss = None
self.trainQPerItem = []
self.trainQ = []
self.validTestQ = []
self.validTestNa = []
self.validTestM = []
self.Pairwise = []
self.Question = []
self.Answer = []
self.Review = []
self.TermtoTermR = []
self.TermtoTermP = []
self.Question_I = []
self.Answer_I = []
self.Review_I = []
self.pairwise = tf.placeholder(dtype = tf.float64, name = 'Pairwise')
self.question = tf.sparse_placeholder(dtype = tf.float64, name = 'Question')
self.answer = tf.sparse_placeholder(dtype = tf.float64, name = 'Answer')
self.review = tf.sparse_placeholder(dtype = tf.float64, name = 'Review')
self.question_I = tf.sparse_placeholder(dtype = tf.float64, name = 'Question')
self.answer_I = tf.sparse_placeholder(dtype = tf.float64, name = 'Answer')
self.review_I = tf.sparse_placeholder(dtype = tf.float64, name = 'Review')
self.termTotermR = tf.sparse_placeholder(dtype = tf.float64, name = 'Review')
self.termTotermP = tf.sparse_placeholder(dtype = tf.float64, name = 'Review')
self.initialize()
self.create_training_data()
self.create_validTest_data()
def initialize(self):
for i in range(self.Items):
self.Pairwise.append([])
self.Question.append([])
self.Answer.append([])
self.Review.append([])
self.TermtoTermR.append([])
self.TermtoTermP.append([])
self.trainQPerItem.append([])
self.Question_I.append([])
self.Answer_I.append([])
self.Review_I.append([])
def create_sparse_one(self, qFeature = None, answer_list = None):
indices = []
values = []
if answer_list is None:
for k, count in sorted(qFeature.items()):
indices.append([0, k])
values.append(count)
if indices == []:
indices.append([0,0])
values.append(np.float64(0))
shape = [1, self.V]
return (np.array(indices), np.array(values), np.array(shape))
else:
for i in range(len(answer_list)):
aFeature = self.corpus.QAnswers[answer_list[i]].aFeature
for k, count in sorted(aFeature.items()):
indices.append([0, i, k])
values.append(count)
if indices == []:
indices.append([0,0,0])
values.append(np.float64(0))
shape = [1, len(answer_list), self.V]
return (np.array(indices), np.array(values), np.array(shape))
def create_sparse_two(self, item, qFeature = None, answer_list = None):
indices = []
values = []
Y = self.corpus.SPerItem[item]
if answer_list is None:
for i in range(len(Y)):
sFeature = self.corpus.Sentences[Y[i]].sFeature
for v1, c1 in sorted(qFeature.items()):
if v1 in sFeature:
indices.append([0, i, v1])
values.append(c1 * sFeature[v1])
if indices == []:
indices.append([0,0,0])
values.append(np.float64(0))
shape = [1, len(Y), self.V]
return (np.array(indices), np.array(values), np.array(shape))
else:
for i in range(len(answer_list)):
aFeature = self.corpus.QAnswers[answer_list[i]].aFeature
for j in range(len(Y)):
sFeature = self.corpus.Sentences[Y[j]].sFeature
for v1, c1 in sorted(aFeature.items()):
if v1 in sFeature:
indices.append([0, i, j, v1])
values.append(c1 * sFeature[v1])
if indices == []:
indices.append([0,0,0,0])
values.append(np.float64(0))
shape = [1, len(answer_list), len(Y), self.V]
return (np.array(indices), np.array(values), np.array(shape))
def create_dense_pairwise(self, item, qId):
Y = self.corpus.SPerItem[item]
Pairwise = np.zeros((1, len(Y), self.PairwiseDim), dtype = np.float64)
for j in range(len(Y)):
Pairwise[0][j] = self.corpus.PairWiseFeature[(qId, Y[j])]
return Pairwise
def create_validTest_data(self):
for i in range(len(self.validTestQ)):
qId = self.validTestQ[i]
item = self.corpus.QAnswers[qId].itemId
question = self.corpus.QAnswers[qId].qFeature
answer_list = [qId, self.validTestNa[i]]
Pairwise = self.create_dense_pairwise(item, qId)
Question = self.create_sparse_one(qFeature = question)
Answer = self.create_sparse_one(answer_list = answer_list)
Review = self.Review[item]
TermtoTermR = self.create_sparse_two(item, qFeature = question)
TermtoTermP = self.create_sparse_two(item, answer_list = answer_list)
Question_I = (Question[0], Question[1] if Question[1].size == 1 and Question[1][0] == 0 else np.full((Question[1].size), 1.0/np.sqrt(Question[1].size)), Question[2])
Answer_I = (Answer[0], Answer[1] if Answer[1].size == 1 and Answer[1][0] == 0 else np.full((Answer[1].size), 1.0/np.sqrt(Answer[1].size)), Answer[2])
Review_I = (Review[0], np.full((Review[1].size), 1.0/np.sqrt(Review[1].size)), Review[2])
self.validTestM.append((Pairwise, Question, Answer, Review, TermtoTermR, TermtoTermP, Question_I, Answer_I, Review_I))
def create_training_data(self):
for i in range(self.Items):
X = self.corpus.QPerItem[i]
for j in range(len(X)):
if j < int(ceil(self.valid* len(X))):
self.trainQPerItem[i].append(X[j])
self.trainQ.append(X[j])
else:
self.validTestQ.append(X[j])
test = int(ceil(len(self.validTestQ) * self.valid_test))
for i in range(len(self.validTestQ)):
if i < test:
na = randint(0, test - 1)
if na == i:
na = (na + 1) % test
else :
na = randint(test, len(self.validTestQ) - 1)
if na == i:
if na == len(self.validTestQ) - 1:
na = test
else:
na = na + 1
self.validTestNa.append(self.validTestQ[na])
for i in range(self.Items):
print "Creating data for ",i
'Calculating Sparse Question Features'
indices = []
values = []
X = self.trainQPerItem[i]
for j in range(int(len(X))):
for k, count in sorted(self.corpus.QAnswers[X[j]].qFeature.items()):
indices.append([j,k])
values.append(count)
shape = [len(X), self.V]
self.Question[i] = (np.array(indices), np.array(values), np.array(shape))
self.Question_I[i] = (np.array(indices), np.full((len(values)), 1.0/np.sqrt(len(values))), np.array(shape))
'Calculating Sparse Answer and Sparse TermtoTermP features'
indices1 = []
values1 = []
indices2 = []
values2 = []
X = self.trainQPerItem[i]
Y = self.corpus.SPerItem[i]
for j in range(len(X)):
for k in range(self.Na):
if k == 0:
aFeature = self.corpus.QAnswers[X[j]].aFeature
else:
na = randint(0, len(self.trainQ) - 1)
if self.trainQ[na] == X[j]:
na = (na + 1) % len(self.trainQ)
aFeature = self.corpus.QAnswers[self.trainQ[na]].aFeature
for l,count in sorted(aFeature.items()):
indices1.append([j,k,l])
values1.append(count)
for m in range(len(Y)):
sFeature = self.corpus.Sentences[Y[m]].sFeature
for v1, c1 in sorted(aFeature.items()):
if v1 in sFeature:
indices2.append([j, k, m, v1])
values2.append(c1 * sFeature[v1])
shape1 = [len(X), self.Na, self.V]
shape2 = [len(X), self.Na, len(Y), self.V]
self.Answer[i] = (np.array(indices1), np.array(values1), np.array(shape1))
self.Answer_I[i] = (np.array(indices1), np.full((len(values1)), 1.0/np.sqrt(len(values1))), np.array(shape1))
self.TermtoTermP[i] = (np.array(indices2), np.array(values2), np.array(shape2))
'Calculating Sparse Review Features at Sentence Level'
indices = []
values = []
X = self.corpus.SPerItem[i]
for j in range(len(X)):
for k, count in sorted(self.corpus.Sentences[X[j]].sFeature.items()):
indices.append([j,k])
values.append(count)
shape = [len(X), self.V]
self.Review[i] = (np.array(indices), np.array(values), np.array(shape))
self.Review_I[i] = (np.array(indices), np.full((len(values)), 1.0/np.sqrt(len(values))), np.array(shape))
'Calculating Dense PairWise and Sparse TermtoTermR features'
X = self.trainQPerItem[i]
Y = self.corpus.SPerItem[i]
pairwise_temp = np.zeros((len(X), len(Y), self.PairwiseDim), dtype = np.float64)
indices = []
values = []
for j in range(len(X)):
for k in range(len(Y)):
pairwise_temp [j][k] = self.corpus.PairWiseFeature[(X[j], Y[k])]
qFeature = self.corpus.QAnswers[X[j]].qFeature
aFeature = self.corpus.Sentences[Y[k]].sFeature
for v1, c1 in sorted(qFeature.items()):
if v1 in aFeature:
indices.append([j, k, v1])
values.append(c1 * aFeature[v1])
shape = [len(X), len(Y), self.V]
self.Pairwise[i] = pairwise_temp
self.TermtoTermR[i] = (np.array(indices), np.array(values), np.array(shape))
def calc_log_loss(self, Pairwise, Question, Answer, Review, TermtoTermR, TermtoTermP, Question_I, Answer_I, Review_I):
#print 'Doing for item %d'%(i)
shape1 = tf.shape(Pairwise)
shape2 = tf.shape(Answer)
nq = shape1[0]
nr = shape1[1]
na = shape2[1]
pairwise = tf.reshape(Pairwise, [-1, self.PairwiseDim])
pairwise = tf.reshape(tf.matmul(pairwise, self.theta), [nq, nr])
termTotermR = tf.sparse_reshape(TermtoTermR, [-1, self.V])
termTotermR = tf.reshape(tf.sparse_tensor_dense_matmul(termTotermR, self.RelvPar), [nq, nr])
QProj = tf.sparse_tensor_dense_matmul(Question_I, self.A)
RProjR = tf.sparse_tensor_dense_matmul(Review_I, self.B)
BilinearR = tf.matmul(QProj, tf.transpose(RProjR))
Relevance = tf.nn.softmax(pairwise + termTotermR + BilinearR)
termTotermP = tf.sparse_reshape(TermtoTermP, [-1, self.V])
termTotermP = tf.reshape(tf.sparse_tensor_dense_matmul(termTotermP, self.PredPar), [nq, na, nr])
AProj = tf.sparse_tensor_dense_matmul(tf.sparse_reshape(Answer_I, [-1, self.V]), self.X)
RProjP = tf.sparse_tensor_dense_matmul(Review_I, self.Y)
BilinearP = tf.reshape(tf.matmul(AProj, tf.transpose(RProjP)), [nq, na, nr])
Prediction = BilinearP + termTotermP
Prediction = tf.expand_dims(Prediction[:,0,:], 1) - Prediction
Prediction = Prediction[:,1:,:]
Prediction = tf.sigmoid(Prediction)
MoE = tf.reduce_sum(tf.multiply(Prediction, tf.expand_dims(Relevance, axis = 1)), axis = 2)
accuracy_count = tf.cast(tf.shape(tf.where(MoE > 0.5))[0], tf.float64)
count = nq * na
log_likelihood = tf.reduce_sum(tf.log(MoE))
R1 = tf.reduce_sum(tf.square(self.A)) + tf.reduce_sum(tf.square(self.B))
R2 = tf.reduce_sum(tf.square(self.X)) + tf.reduce_sum(tf.square(self.Y))
log_likelihood -= self.Lambda * (R1 + R2)
return -1*log_likelihood, MoE, Relevance
def AUC(self, sess):
nq = len(self.validTestQ)
AUC = [0] * nq
AUC_valid = 0
AUC_test = 0
test = int(ceil(len(self.validTestQ) * self.valid_test))
max_na = 1000
for q in range(nq):
print "Calculating AUC for ",q
if q < test:
na_start = 0
na_end = test
else:
na_start = test
na_end = nq
if (na_end - na_start) > max_na:
na_end = na_start + max_na
pairwise, question, answer, review, termtoTermR, termtoTermP, question_I, answer_I, review_I = self.validTestM[q]
itemId = self.corpus.QAnswers[self.validTestQ[q]].itemId
answer_list = self.validTestQ[na_start:na_end]
if self.validTestQ[q] in answer_list:
answer_list.remove(self.validTestQ[q])
answer_list = [self.validTestQ[q]] + answer_list
answer = self.create_sparse_one(answer_list = answer_list)
answer_I = (answer[0], np.full((answer[1].size), 1.0/np.sqrt(answer[1].size)), answer[2])
termtoTermP = self.create_sparse_two(itemId, answer_list = answer_list)
feed_dict = {
self.pairwise : pairwise,
self.question : question,
self.answer : answer,
self.review : review,
self.termTotermR : termtoTermR,
self.termTotermP : termtoTermP,
self.question_I : question_I,
self.answer_I : answer_I,
self.review_I : review_I
}
log_likelihood, MoE, Relevance = sess.run(self.loss, feed_dict = feed_dict)
correct = len(MoE[np.where(MoE > 0.5)])
accuracy = (correct * 1.0) / (len(answer_list) - 1)
if q < test:
AUC_valid += accuracy
else:
AUC_test += accuracy
AUC_valid /= test
AUC_test /= (nq - test)
return AUC_valid, AUC_test
def valid_test_perf(self, sess = None):
test = int(ceil(len(self.validTestQ) * self.valid_test))
MostRelevant = [None] * len(self.validTestQ)
CorrectV = 0
CorrectT = 0
for i in range(len(self.validTestM)):
pairwise, question, answer, review, termtoTermR, termtoTermP, question_I, answer_I, review_I = self.validTestM[i]
feed_dict = {
self.pairwise : pairwise,
self.question : question,
self.answer : answer,
self.review : review,
self.termTotermR : termtoTermR,
self.termTotermP : termtoTermP,
self.question_I : question_I,
self.answer_I : answer_I,
self.review_I : review_I
}
log_likelihood, MoE, Relevance = sess.run(self.loss, feed_dict = feed_dict)
if i < test:
CorrectV += len(MoE[np.where(MoE > 0.5)])
else:
CorrectT += len(MoE[np.where(MoE > 0.5)])
ind = np.argmax(Relevance)
item = self.corpus.QAnswers[self.validTestQ[i]].itemId
sent = self.corpus.SPerItem[item][ind]
MostRelevant[i] = sent
valid_accuracy = (CorrectV * 1.0) / test
test_accuracy = (CorrectT * 1.0) / (len(self.validTestQ) - test)
return valid_accuracy, test_accuracy, MostRelevant
def top_ranked(self, sess, Ktop = 10):
topRanked = []
for i in range(len(self.validTestM)):
h = []
itemId = self.corpus.QAnswers[self.validTestQ[i]].itemId
pairwise, question, answer, review, termtoTermR, termtoTermP, question_I, answer_I, review_I = self.validTestM[i]
feed_dict = {
self.pairwise : pairwise,
self.question : question,
self.answer : answer,
self.review : review,
self.termTotermR : termtoTermR,
self.termTotermP : termtoTermP,
self.question_I : question_I,
self.answer_I : answer_I,
self.review_I : review_I
}
log_likelihood, MoE, Relevance = sess.run(self.loss, feed_dict = feed_dict)
Relevance = Relevance[0]
Relevance = sorted([(Relevance[i], i) for i in range(len(Relevance))])
if len(Relevance) > Ktop:
Relevance = Relevance[len(Relevance)-Ktop:]
for score,ind in Relevance:
sent = self.corpus.SPerItem[itemId][ind]
h.append((score,sent))
topRanked.append(h)
return topRanked
def save_predictions(self, MostRelevant, file):
with open(file, 'w') as file:
maxi = 1000
for q in range(len(self.validTestQ)-1, -1, -1):
ques = self.validTestQ[q]
sent = MostRelevant[q]
json_dump = OrderedDict(
[('itemId', self.corpus.Map.RItemIDMap[self.corpus.QAnswers[ques].itemId]),
('Question', self.corpus.QAnswers[ques].question),
('Answer', self.corpus.QAnswers[ques].answer),
('Review', self.corpus.Sentences[sent].rObj.reviewText),
('Sanitized', self.corpus.Sentences[sent].sent)]
)
json.dump(json_dump, file, indent = 4)
def save_top_ranked(self, topRanked, file):
with open(file, 'w') as file:
for q in range(len(self.validTestQ)):
ques = self.validTestQ[q]
json_dump = OrderedDict(
[('itemId', self.corpus.Map.RItemIDMap[self.corpus.QAnswers[ques].itemId]),
('Question', self.corpus.QAnswers[ques].question),
('Answer', self.corpus.QAnswers[ques].answer)]
)
for j in range(len(topRanked[q])-1 , -1, -1):
score, sent = topRanked[q][j]
sub_dump = OrderedDict(
[('Relevance', score),
('Sentence', self.corpus.Sentences[sent].sent),
('Review', self.corpus.Sentences[sent].rObj.reviewText)]
)
json_dump['Sent'+str(j)] = sub_dump
json.dump(json_dump, file, indent = 4)
def save_model(self, file):
saver = tf.train.Saver()
sess = tf.get_default_session()
saver.save(sess, file)
def restore_model(self, file):
sess = tf.Session()
metafile = file+'.meta'
new_saver = tf.train.import_meta_graph(metafile)
new_saver.restore(sess, tf.train.latest_checkpoint('./'))
all_vars = tf.get_collection('vars')
for v in all_vars:
v_ = sess.run(v)
print(v_)
def train_model(self):
self.theta = tf.Variable(tf.random_uniform([self.PairwiseDim, 1], dtype = tf.float64), name = 'theta' )
self.RelvPar = tf.Variable(tf.random_uniform([self.V, 1], dtype = tf.float64), name = 'RelvPar')
self.A = tf.Variable(tf.random_uniform([self.V, self.rankDim], dtype = tf.float64), name = 'A')
self.B = tf.Variable(tf.random_uniform([self.V, self.rankDim], dtype = tf.float64), name = 'B')
self.PredPar = tf.Variable(tf.random_uniform([self.V, 1], dtype = tf.float64), name = 'PredPar')
self.X = tf.Variable(tf.random_uniform([self.V, self.rankDim], dtype = tf.float64), name = 'X')
self.Y = tf.Variable(tf.random_uniform([self.V, self.rankDim], dtype = tf.float64), name = 'Y')
self.loss = self.calc_log_loss(self.pairwise, self.question, self.answer, self.review, self.termTotermR, self.termTotermP, self.question_I, self.answer_I, self.review_I)
train_step = tf.train.AdamOptimizer(1e-2).minimize(self.loss[0])
sess = tf.Session()
sess.run(tf.global_variables_initializer())
for i in range(self.numIter):
log_likelihood = 0.0
accuracy_count = 0
count = 0
for j in range(self.Items):
feed_dict = {
self.pairwise : self.Pairwise[j],
self.question : self.Question[j],
self.answer : self.Answer[j],
self.review : self.Review[j],
self.termTotermR : self.TermtoTermR[j],
self.termTotermP : self.TermtoTermP[j],
self.question_I : self.Question_I[j],
self.answer_I : self.Answer_I[j],
self.review_I : self.Review_I[j]
}
train, result = sess.run([train_step, self.loss], feed_dict = feed_dict)
log_likelihood += result[0]
accuracy_count += len(result[1][np.where(result[1] > 0.5)])
count += result[1].size
accuracy = (accuracy_count * 1.0) / count
valid, test, topRanked = self.valid_test_perf(sess)
print "For Training Epoch ", i
print "--------------------------------------"
print "Training Data:"
print "\tlog_likelihood: ", log_likelihood
print "\taccuracy: ", accuracy
print "Valid Data:"
print "\taccuracy: ", valid
print "Test Data:"
print "\taccuracy: ", test
print "--------------------------------------"
return sess