-
Notifications
You must be signed in to change notification settings - Fork 2
/
Neuron-embedding.py
151 lines (111 loc) · 4.47 KB
/
Neuron-embedding.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
#!/usr/bin/env python2.7
from __future__ import print_function
import numpy as np
import json
import sys
import theano
import pickle
from keras.preprocessing.sequence import pad_sequences
from keras.models import Sequential, Graph
from keras.layers.core import Dense, TimeDistributedDense, Activation, Dropout, AutoEncoder, Flatten, Merge
from keras.layers.embeddings import Embedding
from keras.layers import containers
from keras.layers.recurrent import SimpleRNN, LSTM
from keras.utils import generic_utils
from sklearn.feature_extraction.text import CountVectorizer
from sklearn.metrics import accuracy_score
batch_size = 20
n_d = 100
if __name__ == "__main__":
if len(sys.argv) < 3:
print("Usage: {} [training reactions json] [chemical embeddings]".format(sys.argv[0]))
quit(1)
# Read training data
training_data_fn = sys.argv[1]
print("Reading input...", end=" ")
with open(training_data_fn) as f:
training_data = json.load(f)
print("done.")
# Read embeddings
embeddings_fn = sys.argv[2]
print("Reading embeddings...", end=" ")
with open(embeddings_fn) as f:
embeddings = pickle.load(f)
print("done.")
# Fit the countvectorizer to the corpus
token_regex = r"(?u)([\(\)\[\]]|\b\w+\b)"
cv = CountVectorizer(token_pattern=token_regex, min_df=1)
an = cv.build_analyzer()
print("Training length: {}, embedding vector width: {}, batch_size: {}".format(len(training_data), n_d, batch_size))
def word2embedding(w):
tokens = an(w.strip())
word_embeddings = np.zeros((batch_size, n_d), dtype=theano.config.floatX)
# (pre-)pad out tokens
if len(tokens) < batch_size:
tokens = [""] * (batch_size - len(tokens)) + tokens
for i, token in enumerate(tokens[:batch_size]):
if token in embeddings:
word_embeddings[i] = embeddings[token]
return word_embeddings
# Build the training vectors
X = np.zeros((len(training_data), batch_size, 2*n_d), dtype=theano.config.floatX)
Y = np.zeros((len(training_data)), dtype=np.bool)
np.random.shuffle(training_data)
for i, (c1, c2, reacts) in enumerate(training_data):
X[i] = np.concatenate((word2embedding(c1), word2embedding(c2)), axis=1)
Y[i] = reacts
# Create training/development split
split = int(len(training_data) * 0.9)
X_train, X_dev = X[:split], X[split:]
Y_train, Y_dev = Y[:split], Y[split:]
def generate_batches(number):
count = 0
while count < X_train.shape[0]:
yield X_train[count:count+number], Y_train[count:count+number].reshape((-1, 1))
count += number
print("Constructing model...", end=" ")
dropout = 0.5
model = Graph()
model = Sequential()
model.add(LSTM(output_dim=2*n_d, input_shape=(batch_size, 2*n_d), activation="tanh", return_sequences=False))
model.add(Dropout(dropout))
model.add(Dense(1, activation="sigmoid"))
model.compile(loss="binary_crossentropy", optimizer="adam", class_mode="binary")
print("done.")
ts = []
ds = []
data_file = "performance/data.tsv"
def score(iteration):
train_preds = model.predict_classes(X_train)
dev_preds = model.predict_classes(X_dev)
train_acc = accuracy_score(train_preds, Y_train)
dev_acc = accuracy_score(dev_preds, Y_dev)
print("Train accuracy: {:.4f}".format(train_acc))
print("Dev accuracy : {:.4f}".format(dev_acc))
ts.append(train_acc)
ds.append(dev_acc)
with open(data_file, "a+") as f:
f.write("{}\t{}\t{}\n".format(iteration, train_acc, dev_acc))
# Train the model
print("Training...")
with open(data_file, "w") as f:
f.write("iteration\ttraining\tdevelopment\n")
iterations = 200
nb_epochs = 10
try:
for iteration in xrange(iterations):
print("ITERATION", iteration)
for epoch in xrange(nb_epochs):
print("Epoch", epoch)
progbar = generic_utils.Progbar(len(X_train))
for x_train, y_train in generate_batches(128):
loss = model.train_on_batch(x_train, y_train)
progbar.add(x_train.shape[0], values=[("train loss", loss[0])])
# Score the model
score(iteration)
except KeyboardInterrupt:
pass
print("\ndone.")
score(iteration)
print("Previous training scores:", ts)
print("Previous development scores:", ds)