DECODA_binary_BOW_SPE.py
2.94 KB
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
# coding: utf-8
# In[2]:
# Import
import pandas
# Alignement
from alignment.sequence import Sequence
from alignment.vocabulary import Vocabulary
from alignment.sequencealigner import *
import nltk
import codecs
import gensim
from scipy import sparse
import itertools
from sklearn.feature_extraction.text import CountVectorizer
import scipy.sparse
import scipy.io
from sklearn import preprocessing
from keras.models import Sequential
from keras.layers.core import Dense, Dropout, Activation,AutoEncoder
from keras.optimizers import SGD,Adam
from keras.layers import containers
from mlp import *
import mlp
import sklearn.metrics
import shelve
import pickle
from utils import *
import sys
# In[4]:
db=shelve.open("DECODA_sparse.shelve")
out_db=shelve.open("{}.shelve".format(sys.argv[1]),writeback=True)
# In[6]:
ASR_sparse=db["ASR_SPARSE"]
TRS_sparse=db["TRS_SPARSE"]
LABEL=db["LABEL"]
def select(elm):
return int(elm.split("_")[-1])
print "gogo autoencoder ASR"
hidden_size=3076
hidden_size2=2048
input_activation="relu"
out_activation="relu"
loss="mse"
epochs=500
patience=50
sgd = Adam(lr=0.001) #SGD(lr=0.05,momentum=0.9)
autoencode=Sequential()
autoencode.add(Dense(ASR_sparse["TRAIN"].shape[1],hidden_size,init='glorot_uniform',activation=input_activation))
autoencode.add(Dense(hidden_size,hidden_size2,init='glorot_uniform',activation=input_activation))
autoencode.add(Dense(hidden_size2,hidden_size,init="glorot_uniform",activation=out_activation))
autoencode.add(Dense(hidden_size,ASR_sparse["DEV"].todense().shape[1],init="glorot_uniform",activation=out_activation))
autoencode.compile(optimizer=sgd,loss=loss)
# In[ ]:
autoencode.fit(ASR_sparse["TRAIN"].todense(),TRS_sparse["TRAIN"].todense(),nb_epoch=epochs,batch_size=16,
callbacks=[keras.callbacks.EarlyStopping(monitor='val_loss',
patience=patience, verbose=0)], validation_data=(ASR_sparse["DEV"].todense(),TRS_sparse["DEV"].todense()),verbose=1)
# In[ ]:
auto_decoder=Sequential()
auto_decoder.add(Dense(ASR_sparse["DEV"].todense().shape[1],hidden_size,init='glorot_uniform',activation=input_activation,weights=autoencode.get_weights()[:2]))
auto_decoder.add(Dense(hidden_size,hidden_size2,init='glorot_uniform',activation=input_activation,weights=autoencode.get_weights()[2:4]))
auto_decoder.compile(optimizer=sgd,loss=loss)
# In[77]:
#autoencode.predict(ASR_sparse["DEV"].todense())
# In[ ]:
print "auto encoder et auto decoder asr okay"
ASR_sparse_AE={}
TRS_sparse_AE={}
for i in ASR_sparse.keys():
ASR_sparse_AE[i]=auto_decoder.predict(ASR_sparse[i].todense())
TRS_sparse_AE[i]=auto_decoder.predict(TRS_sparse[i].todense())
#TRS_sparse[i]=dico.transform(TRS[i][2])
out_db["ASR_SPELIKE_H2_RELU"]=ASR_sparse_AE
out_db["TRS_SPELIKE_H2_RELU"]=TRS_sparse_AE
ASR_AE={}
for i in TRS_sparse.keys():
ASR_AE[i]=autoencode.predict(ASR_sparse[i].todense())
out_db["ASR_SPELIKE_OUT_RELU"]=ASR_AE
db.close()
out_db.close()