-
Notifications
You must be signed in to change notification settings - Fork 7
/
tagcn.py
166 lines (117 loc) · 5.85 KB
/
tagcn.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
import tensorflow as tf
import numpy as np
from utils import *
from metrics import *
from inits import *
flags = tf.app.flags
FLAGS = flags.FLAGS
flags.DEFINE_string('dataset', 'cora', 'Dataset string.') # 'cora', 'citeseer', 'pubmed'
flags.DEFINE_string('model', 'gcn', 'Model string.') # 'gcn', 'gcn_cheby', 'dense'
flags.DEFINE_float('learning_rate', 0.01, 'Initial learning rate.')
flags.DEFINE_integer('epochs', 200, 'Number of epochs to train.')
flags.DEFINE_integer('hidden1', 16, 'Number of units in hidden layer 1.')
flags.DEFINE_float('dropout', 0.5, 'Dropout rate (1 - keep probability).')
flags.DEFINE_float('weight_decay', 5e-4, 'Weight for L2 loss on embedding matrix.')
flags.DEFINE_integer('early_stopping', 10, 'Tolerance for early stopping (# of epochs).')
flags.DEFINE_integer('max_degree', 3, 'Maximum Chebyshev polynomial degree.')
adj, features, y_train, y_val, y_test, train_mask, val_mask, test_mask = load_data(FLAGS.dataset)
features = features.todense()
# g = tf.get_default_graph()
# [op.name for op in g.get_operations()]
path_weight_matrix = np.load("path_weights.dat")
path_weight_matrix = path_weight_matrix.astype('float32')
from inits import *
# var_gs = {}
name = '01'
Fl = 8
Kl = 2
Cl = features.shape[1]
Nl = features.shape[0]
from tensorflow.python.framework.ops import reset_default_graph
reset_default_graph()
# define weights, biases
vars_F={}
with tf.variable_scope(name + '_vars'):
for k in range(Kl):
vars_F['weights_' + str(0) + '_' + str(k)] = glorot([Cl,Fl],name=('weights_' + str(0) + '_' + str(k)))
# tf.get_variable(name=('weights_' + str(0) + '_' + str(k)),
# initializer=tf.contrib.layers.xavier_initializer(),
# shape=[Cl,Fl])
initial = tf.zeros([Nl,Fl], dtype=tf.float32)
vars_F['bias_' + str(0)] = tf.Variable(initial, name='bias')
for k in range(Kl):
vars_F['weights_' + str(1) + '_' + str(k)] = glorot([Fl,7],name=('weights_' + str(1) + '_' + str(k)))
# tf.get_variable(name=('weights_' + str(1) + '_' + str(k)),
# initializer=tf.contrib.layers.xavier_initializer(),
# shape=[Fl,7]) # prev [Fl,Fl]
initial = tf.zeros([Nl,7], dtype=tf.float32) # prev [Nl,Fl]
vars_F['bias_' + str(1)] = tf.Variable(initial, name='bias')
features_m = tf.placeholder(tf.float32, shape=[features.shape[0], features.shape[1]])
pwm = tf.placeholder(tf.float32, [Nl, Nl, Kl])
outputs=[]
def layer(input_t,output_dim,num):
# droupout
input_t = tf.nn.dropout(input_t, 1-FLAGS.dropout)
conv = np.zeros(output_dim,dtype=np.float32)
for k in range(Kl):
w_k = pwm[:,:,k]
s = tf.matmul(w_k,input_t)
G_k = vars_F['weights_' + str(num) + '_' + str(k)]
res = tf.matmul(s,G_k)
conv = tf.add(conv,res)
print(conv.shape)
# add bias
bias = vars_F['bias_' + str(num)]
conv = tf.add(conv,bias)
# apply non-linearity to first layer
if num == 0:
conv = tf.nn.relu(conv)
return conv
res = layer(features_m,output_dim=[Nl,Fl],num=0)
conv = layer(res,output_dim=[Nl,7],num=1)
# for training
accuracy1 = masked_accuracy(conv,y_train, train_mask)
loss1= 0
# weight decay loss
loss1 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(0) + '_' + str(0)])
loss1 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(0) + '_' + str(1)])
loss1 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['bias_' + str(0)])
loss1 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(1) + '_' + str(0)])
loss1 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(1) + '_' + str(1)])
loss1 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['bias_' + str(1)])
# Cross entropy error
loss1 += masked_softmax_cross_entropy(conv, y_train, train_mask)
optimizer = tf.train.AdamOptimizer(learning_rate=FLAGS.learning_rate)
opt_op = optimizer.minimize(loss1)
# for testing
accuracy2 = masked_accuracy(conv,y_test, test_mask)
loss2= 0
# weight decay loss
loss2 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(0) + '_' + str(0)])
loss2 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(0) + '_' + str(1)])
loss2 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['bias_' + str(0)])
loss2 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(1) + '_' + str(0)])
loss2 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(1) + '_' + str(1)])
loss2 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['bias_' + str(1)])
# Cross entropy error
loss2 += masked_softmax_cross_entropy(conv, y_test, test_mask)
# for validation
accuracy3 = masked_accuracy(conv,y_val, val_mask)
loss3= 0
# weight decay loss
loss3 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(0) + '_' + str(0)])
loss3 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(0) + '_' + str(1)])
loss3 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['bias_' + str(0)])
loss3 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(1) + '_' + str(0)])
loss3 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['weights_' + str(1) + '_' + str(1)])
loss3 += FLAGS.weight_decay * tf.nn.l2_loss(vars_F['bias_' + str(1)])
# Cross entropy error
loss3 += masked_softmax_cross_entropy(conv, y_val, val_mask)
sess = tf.Session()
sess.run(tf.global_variables_initializer())
for epoch in range(FLAGS.epochs):
evals = sess.run([opt_op,loss1,accuracy1],feed_dict={features_m:features,pwm:path_weight_matrix})
evals2 = sess.run([loss3,accuracy3],feed_dict={features_m:features,pwm:path_weight_matrix})
print("Test",evals[1], evals[2],"Validation",evals2[0], evals2[1])
outs_val = sess.run([loss2, accuracy2], feed_dict={features_m:features,pwm:path_weight_matrix})
print(outs_val[0], outs_val[1])