-
Notifications
You must be signed in to change notification settings - Fork 1
/
tf_samplergcn.py
239 lines (182 loc) · 9.93 KB
/
tf_samplergcn.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
#利用图卷积进行图节点分类,但在采样邻居节点时只选取部分邻居
# coding=utf-8
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
import tf_geometric as tfg
import tensorflow as tf
from tensorflow import keras
from tf_geometric.utils.graph_utils import add_self_loop_edge
from tf_geometric.utils.graph_utils import RandomNeighborSampler#邻居随机采样
graph, (train_index, valid_index, test_index) = tfg.datasets.CoraDataset().load_data()
print(graph.x.shape, graph.edge_index.shape)#(2708, 1433) (2, 10556)
num_classes = graph.y.max() + 1#7
drop_rate = 0.2
neighbor_sampler = RandomNeighborSampler(graph.edge_index)
#邻居节点随机采样，它有两个参数，一个是k,规定采样邻居个数，若邻居个数不足k个的，则对邻居进行多次采样直至k个样本
#另一个参数是ratio,它控制采样邻居的比例，两个参数只需提供其中一个
#两层GCNloss为0.45，acc为0.81
class GCN(tf.keras.Model):#对照组，看一个4层GCN的训练效果, loss = 1.1754034757614136 accuracy = 0.503000020980835
def __init__(self, rate, num_classes):
super(GCN, self).__init__()
self.rate = rate
self.num_classes = num_classes
self.gcn0 = tfg.layers.GCN(256, activation = tf.nn.relu)
self.gcn1 = tfg.layers.GCN(128, activation=tf.nn.relu)#attention_units * num_heads = 维度（64）
self.gcn2 = tfg.layers.GCN(64, activation=tf.nn.relu)
self.gcn3 = tfg.layers.GCN(self.num_classes)
self.dropout = keras.layers.Dropout(self.rate)
def call(self, graph):
h = graph.x
h = self.dropout(h)
h = self.gcn0([h, graph.edge_index, graph.edge_weight], cache=graph.cache)#GCN操作格式
h = self.dropout(h)
h = self.gcn1([h, graph.edge_index, graph.edge_weight], cache=graph.cache)
h = self.dropout(h)
h = self.gcn2([h, graph.edge_index, graph.edge_weight], cache=graph.cache)
h = self.dropout(h)
h = self.gcn3([h, graph.edge_index, graph.edge_weight], cache=graph.cache)
return h
class KSamplerGCN(tf.keras.Model):#一个指定邻居个数的采样GCN(4层升采样),loss = 0.4289660155773163 accuracy = 0.6480000019073486
def __init__(self, rate, num_classes):
super(KSamplerGCN, self).__init__()
self.rate = rate
self.num_classes = num_classes
self.gcn0 = tfg.layers.GCN(256, activation = tf.nn.relu)
self.gcn1 = tfg.layers.GCN(128, activation=tf.nn.relu)
self.gcn2 = tfg.layers.GCN(64, activation=tf.nn.relu)
self.gcn3 = tfg.layers.GCN(self.num_classes)
self.dropout = keras.layers.Dropout(self.rate)
def call(self, graph):
h = graph.x
h = self.dropout(h)
sampled_edge_index_0, sampled_edge_weight_0 = neighbor_sampler.sample(k=6)#第一层升采样
h = self.gcn0([h, sampled_edge_index_0, sampled_edge_weight_0], cache=graph.cache)
h = self.dropout(h)
sampled_edge_index_1, sampled_edge_weight_1 = neighbor_sampler.sample(k=6)
h = self.gcn1([h, sampled_edge_index_1, sampled_edge_weight_1], cache=graph.cache)
h = self.dropout(h)
sampled_edge_index_2, sampled_edge_weight_2 = neighbor_sampler.sample(k=5)
h = self.gcn2([h, sampled_edge_index_2, sampled_edge_weight_2], cache=graph.cache)
h = self.dropout(h)
sampled_edge_index_3, sampled_edge_weight_3 = neighbor_sampler.sample(k=5)
h = self.gcn3([h, sampled_edge_index_3, sampled_edge_weight_3], cache=graph.cache)
return h
class KRandSamplerGCN(tf.keras.Model):#一个指定邻居个数的采样GCN.采样时采用随机采样技术，但是每个节点的采样数都为图中节点的平均邻居数
#这样对邻居多的节点限制聚合，对邻居少的节点突出邻居特征，达到loss=0.2, acc=0.75，效果最好
def __init__(self, rate, num_classes):
super(KRandSamplerGCN, self).__init__()
self.rate = rate
self.num_classes = num_classes
self.gcn0 = tfg.layers.GCN(256, activation = tf.nn.relu)
self.gcn1 = tfg.layers.GCN(128, activation=tf.nn.relu)
self.gcn2 = tfg.layers.GCN(64, activation=tf.nn.relu)
self.gcn3 = tfg.layers.GCN(self.num_classes)
self.dropout = keras.layers.Dropout(self.rate)
def call(self, graph):
h = graph.x
h = self.dropout(h)
sampled_edge_index_0, sampled_edge_weight_0 = neighbor_sampler.sample(k=4)
h = self.gcn0([h, sampled_edge_index_0, sampled_edge_weight_0], cache=graph.cache)
h = self.dropout(h)
sampled_edge_index_1, sampled_edge_weight_1 = neighbor_sampler.sample(k=4)
h = self.gcn1([h, sampled_edge_index_1, sampled_edge_weight_1], cache=graph.cache)
h = self.dropout(h)
sampled_edge_index_2, sampled_edge_weight_2 = neighbor_sampler.sample(k=4)
h = self.gcn2([h, sampled_edge_index_2, sampled_edge_weight_2], cache=graph.cache)
h = self.dropout(h)
sampled_edge_index_3, sampled_edge_weight_3 = neighbor_sampler.sample(k=4)
h = self.gcn3([h, sampled_edge_index_3, sampled_edge_weight_3], cache=graph.cache)
return h
class RatioSamplerGCN(tf.keras.Model):#一个指定采样比例的邻居降采样GCN,loss = 0.4289924204349518 accuracy = 0.6389999985694885
def __init__(self, rate, num_classes):
super(RatioSamplerGCN, self).__init__()
self.rate = rate
self.num_classes = num_classes
self.gcn0 = tfg.layers.GCN(256, activation = tf.nn.relu)
self.gcn1 = tfg.layers.GCN(128, activation=tf.nn.relu)）
self.gcn2 = tfg.layers.GCN(64, activation=tf.nn.relu)
self.gcn3 = tfg.layers.GCN(self.num_classes)
self.dropout = keras.layers.Dropout(self.rate)
def call(self, graph):
h = graph.x
h = self.dropout(h)
sampled_edge_index_0, sampled_edge_weight_0 = neighbor_sampler.sample(ratio=1)
h = self.gcn0([h, sampled_edge_index_0, sampled_edge_weight_0], cache=graph.cache)
h = self.dropout(h)
sampled_edge_index_1, sampled_edge_weight_1 = neighbor_sampler.sample(ratio=0.8)
h = self.gcn1([h, sampled_edge_index_1, sampled_edge_weight_1], cache=graph.cache)
h = self.dropout(h)
sampled_edge_index_2, sampled_edge_weight_2 = neighbor_sampler.sample(ratio=0.6)
h = self.gcn2([h, sampled_edge_index_2], cache=graph.cache)
h = self.gcn2([h, sampled_edge_index_2], cache=graph.cache)
h = self.dropout(h)
sampled_edge_index_3, sampled_edge_weight_3 = neighbor_sampler.sample(ratio=0.5)
h = self.gcn3([h, sampled_edge_index_3, sampled_edge_weight_3], cache=graph.cache)
return h
def compute_loss(logits, mask_index, vars):
masked_logits = tf.gather(logits, mask_index)#获取所求元素在集中的位置
masked_labels = tf.gather(graph.y, mask_index)
losses = tf.nn.softmax_cross_entropy_with_logits(
logits=masked_logits,
labels=tf.one_hot(masked_labels, depth=num_classes)
)
kernel_vals = [var for var in vars if "kernel" in var.name]
l2_losses = [tf.nn.l2_loss(kernel_var) for kernel_var in kernel_vals]#添加l2损失
return tf.reduce_mean(losses) + tf.add_n(l2_losses) * 5e-4
def evaluate(logits):#评估
masked_logits = tf.gather(logits, test_index)#获取测试集对应元素
masked_labels = tf.gather(graph.y, test_index)
y_pred = tf.argmax(masked_logits, axis=-1, output_type=tf.int32)
accuracy_m = keras.metrics.Accuracy()
accuracy_m.update_state(masked_labels, y_pred)
return accuracy_m.result().numpy()
optimizer = tf.keras.optimizers.Adam(learning_rate=5e-3)
kgcn = KSamplerGCN(drop_rate, num_classes)
rgcn = RatioSamplerGCN(drop_rate, num_classes)
krgcn = KRandSamplerGCN(drop_rate, num_classes)
gcn = GCN(drop_rate, num_classes)
def train_step_gcn():
for step in range(800):
with tf.GradientTape() as tape:
logits = gcn.call(graph)
loss = compute_loss(logits, train_index, tape.watched_variables())
vars = gcn.trainable_variables
grads = tape.gradient(loss, vars)
optimizer.apply_gradients(zip(grads, vars))
if step % 20 == 0:
accuracy = evaluate(logits)
print("step = {}\tloss = {}\taccuracy = {}".format(step, loss, accuracy))
def train_step_kgcn():
for step in range(800):
with tf.GradientTape() as tape:
logits = kgcn.call(graph)
loss = compute_loss(logits, train_index, tape.watched_variables())
vars = kgcn.trainable_variables
grads = tape.gradient(loss, vars)
optimizer.apply_gradients(zip(grads, vars))
if step % 20 == 0:
accuracy = evaluate(logits)
print("step = {}\tloss = {}\taccuracy = {}".format(step, loss, accuracy))
def train_step_krgcn():
for step in range(800):
with tf.GradientTape() as tape:
logits = krgcn.call(graph)
loss = compute_loss(logits, train_index, tape.watched_variables())
vars = krgcn.trainable_variables
grads = tape.gradient(loss, vars)
optimizer.apply_gradients(zip(grads, vars))
if step % 20 == 0:
accuracy = evaluate(logits)
print("step = {}\tloss = {}\taccuracy = {}".format(step, loss, accuracy))
def train_step_rgcn():
for step in range(800):
with tf.GradientTape() as tape:
logits = rgcn.call(graph)
loss = compute_loss(logits, train_index, tape.watched_variables())
vars = rgcn.trainable_variables
grads = tape.gradient(loss, vars)
optimizer.apply_gradients(zip(grads, vars))
if step % 20 == 0:
accuracy = evaluate(logits)
print("step = {}\tloss = {}\taccuracy = {}".format(step, loss, accuracy))
train_step_rgcn()