-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathutils.py
121 lines (90 loc) · 3.48 KB
/
utils.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
import numpy as np
import scipy.sparse as sp
import networkx as nx
import pickle as pkl
import tensorflow as tf
def preprocess_features(features):
row_sum = np.array(features.sum(1))
reverse_row_sum = np.power(row_sum,-1).flatten()
reverse_row_sum[np.isinf(reverse_row_sum)] = 0.
new_features = sp.diags(reverse_row_sum).dot(features)
return new_features
def load_data(datasetname):
names = ['x','tx','allx','y','ty','ally','graph']
objects = {}
for name in names:
with open("data/ind.{}.{}".format(datasetname, name),'rb') as f:
objects[name] = pkl.load(f, encoding='latin1')
with open("data/ind.{}.test.index".format(datasetname), 'r') as f:
test_index = []
for line in f.readlines():
test_index.append(int(line.strip()))
test_index_reorder = np.sort(test_index)
whole_features = sp.vstack((objects['allx'], objects['tx'])).tolil()
whole_features[test_index] = whole_features[test_index_reorder]
num_nodes = whole_features.shape[0]
adj = nx.adjacency_matrix(nx.from_dict_of_lists(objects['graph']))
whole_labels = np.r_[objects['ally'], objects['ty']]
whole_labels[test_index] = whole_labels[test_index_reorder]
train_idx = np.arange(len(objects['y']))
val_idx = np.arange(len(objects['y']), len(objects['y'])+ 500)
test_idx = test_index_reorder
return adj, whole_features, whole_labels, train_idx, val_idx, test_idx
def sample_mask(idx, l):
mask = np.zeros(l)
mask[idx] = 1
return np.array(mask, dtype=np.bool)
def generate_mask_data(node_labels, idx):
"""
idx: train_idx / val_idx / test_idx
return:
dataset_labels: y_train / y_val / y_test
dataset_mask: train_mask / val_mask / test_mask
"""
dataset_labels = np.zeros(node_labels.shape)
dataset_mask = sample_mask(idx, node_labels.shape[0])
dataset_labels[dataset_mask] = node_labels[dataset_mask]
return dataset_labels, dataset_mask
def scatter_sum(updates, indices, N):
"""
tf.unsorted_segment_sum( data, segment_ids, num_segments, name=None )
data = [
[1,2,3],[4,5,6],[7,8,9]
]
segment_ids = [2,0,1]
num_segment = 3
output = unsorted_segment_sum(data, segment_ids, num_segmentss)
output[2] = data[0]
output[0] = data[1]
output[1] = data[2]
若segment_ids = [2,0,0]
则output[2] = data[0]
output[1] = [0,0,0]
output[0] = data[1]+ data[2]
updates： neighbor_features
indices: center_node_ids
N: N segments
indices = [0,0,0,1,1,.....,2077,2077,2078,2078,2078]
output[0] = updates[0] + updates[1] + updates[2]
output[1] = updates[3] + updates[4] +..
...
output[2078] = update[...] + update[...] + ...
"""
return tf.math.unsorted_segment_sum(updates, indices, N)
def scatter_mean(updates, indices, N):
return tf.math.unsorted_segment_mean(updates, indices, N)
def scatter_max(updates, indices, N):
return tf.math.unsorted_segment_max(updates, indices, N)
def scatter_min(updates, indices, N):
return tf.math.unsorted_segment_min(updates, indices , N)
def scatter_prod(updates, indices, N):
return tf.math.unsorted_segment_prod(updates, indices, N)
def convert_csr_to_SparseTensor(csr_matrix):
if sp.isspmatrix_csr(csr_matrix):
csr_matrix = csr_matrix.tocoo()
row = csr_matrix.row
col = csr_matrix.col
pos = np.c_[row,col]
data = csr_matrix.data
shape = csr_matrix.shape
return tf.SparseTensor(pos, data, shape)