-
Notifications
You must be signed in to change notification settings - Fork 2
/
Copy pathdata_preparation.py
132 lines (101 loc) · 6.45 KB
/
data_preparation.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
import numpy as np
class Data():
def __init__(self, args):
self.parse_args(args)
self.load_data()
def parse_args(self, args):
self.dataset_name = args.dataset_name
self.x = args.x
self.trans_induc = args.trans_induc
self.minibatch_size = args.minibatch_size
if self.trans_induc == 'transductive':
self.training_ratio = 1
elif self.trans_induc == 'inductive':
self.training_ratio = args.training_ratio - args.training_ratio * 0.1
self.validation_ratio = args.training_ratio * 0.1
self.test_ratio = 1 - args.training_ratio
def load_data(self):
self.doc = self.doc_preprocessing(np.loadtxt('./cora/' + self.dataset_name + '/content.txt'))
self.num_doc = len(self.doc)
self.label = np.loadtxt('./cora/' + self.dataset_name + '/label.txt')
self.adjacency_matrix = self.generate_symmetric_adjacency_matrix(np.loadtxt('./cora/' + self.dataset_name + '/adjacency_matrix.txt'))
self.links = self.generate_links(self.adjacency_matrix)
np.random.shuffle(self.links)
self.voc = np.genfromtxt('./cora/' + self.dataset_name + '/voc.txt', dtype=str)
self.num_tokens = len(self.voc)
if self.trans_induc == 'transductive':
self.doc_training, self.doc_test = self.doc, self.doc
self.label_training, self.label_test = self.label, self.label
self.adjacency_matrix_training, self.adjacency_matrix_test = self.adjacency_matrix, self.adjacency_matrix
self.links_training = self.links
elif self.trans_induc == 'inductive':
self.doc_training, self.doc_test = self.doc[:int(self.num_doc * self.training_ratio)], self.doc[int(self.num_doc * (self.training_ratio + self.validation_ratio)):]
self.label_training, self.label_test = self.label[:int(self.num_doc * self.training_ratio)], self.label[int(self.num_doc * (self.training_ratio + self.validation_ratio)):]
self.adjacency_matrix_training, self.adjacency_matrix_test = \
self.adjacency_matrix[:int(self.num_doc * self.training_ratio), :int(self.num_doc * self.training_ratio)], \
self.adjacency_matrix[int(self.num_doc * (self.training_ratio + self.validation_ratio)):, :int(self.num_doc * self.training_ratio)]
self.links_training, self.links_test = self.split_links(self.links)
self.num_minibatch = int(np.ceil(len(self.links_training) / self.minibatch_size))
self.minibatch_data = {}
for minibatch_index in range(1, self.num_minibatch + 1):
self.prepare_minibatch(num_minibatch=self.num_minibatch, minibatch_index=minibatch_index)
self.minibatch_data[minibatch_index] = {}
self.minibatch_data[minibatch_index]['sampling_links'] = self.sampling_links
self.minibatch_data[minibatch_index]['neighbor_ids'] = self.neighbor_ids
self.minibatch_data[minibatch_index]['segment_ids'] = self.segment_ids
if self.x == 0:
self.input_training = self.doc_training
self.input_test = self.doc_test
elif self.x == 1:
self.input_training = np.concatenate([self.doc_training, self.adjacency_matrix_training], axis=1)
self.input_test = np.concatenate([self.doc_test, self.adjacency_matrix_test], axis=1)
def doc_preprocessing(self, doc):
doc_preprocessed = []
for row in doc:
max_row = np.log(1 + np.max(row))
doc_preprocessed.append(np.log(1 + row) / max_row)
return np.asarray(doc_preprocessed)
def generate_symmetric_adjacency_matrix(self, adjacency_matrix):
adjacency_matrix_symm = np.zeros((len(adjacency_matrix), len(adjacency_matrix)))
for row_idx in range(len(adjacency_matrix)):
for col_idx in range(len(adjacency_matrix)):
if adjacency_matrix[row_idx, col_idx] == 1:
adjacency_matrix_symm[row_idx, col_idx] = 1
adjacency_matrix_symm[col_idx, row_idx] = 1
if row_idx == col_idx:
adjacency_matrix_symm[row_idx, col_idx] = 1
return adjacency_matrix_symm
def generate_links(self, adjacency_matrix):
links = []
for row_idx in range(len(adjacency_matrix)):
for col_idx in range(len(adjacency_matrix)):
if adjacency_matrix[row_idx, col_idx] == 1:
links.append([row_idx, col_idx])
return np.asarray(links)
def split_links(self, links):
links_training, links_test = [], []
for link in links:
if link[0] < int(self.num_doc * self.training_ratio) and link[1] < int(self.num_doc * self.training_ratio):
links_training.append([link[0], link[1]])
elif link[0] >= int(self.num_doc * (self.training_ratio + self.validation_ratio)) and link[1] >= int(self.num_doc * (self.training_ratio + self.validation_ratio)):
links_test.append([link[0], link[1]])
return np.asarray(links_training), np.asarray(links_test)
def prepare_minibatch(self, num_minibatch, minibatch_index):
self.sampling_links = self.sample_minibatch_links(num_minibatch, minibatch_index)
self.neighbor_ids, self.segment_ids = self.prepare_neighbors()
def sample_minibatch_links(self, num_minibatch, minibatch_index):
if minibatch_index == num_minibatch:
sampling_links = self.links_training[self.minibatch_size * (minibatch_index - 1):]
if self.minibatch_size - len(sampling_links) != 0:
indices = np.random.choice(len(self.links_training), self.minibatch_size - len(sampling_links), replace=False)
sampling_links = np.concatenate((sampling_links, self.links_training[indices]), axis=0)
else:
sampling_links = self.links_training[self.minibatch_size * (minibatch_index - 1):self.minibatch_size * minibatch_index]
return sampling_links
def prepare_neighbors(self):
neighbor_ids, segment_ids = [], []
for idx, link in enumerate(self.sampling_links):
neighbor_ids_tmp = self.links_training[self.links_training[:, 0] == link[0]][:, 1]
neighbor_ids.extend(neighbor_ids_tmp)
segment_ids += len(neighbor_ids_tmp) * [idx]
return np.asarray(neighbor_ids), np.asarray(segment_ids)