/
graph.py
199 lines (168 loc) · 6.42 KB
/
graph.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
import numpy as np
from scipy.sparse import coo_matrix
import scipy.sparse as sp
from utils import read_csv, Vocab, read_txt
import collections
from pathlib import Path
import typing
class StaticGraph:
def __init__(self):
self._adj = None
self._adj_csr = None
self._vocab = None
self._label_vocab = None
self._node_label = None
self._node_feature = None
@property
def adj(self) -> sp.coo_matrix:
return self._adj
@property
def adj_csr(self) -> sp.csr_matrix:
return self._adj_csr
@adj.setter
def adj(self, x):
self._adj = x
self._adj_csr = self._adj.tocsr()
@property
def node_size(self) -> int:
return self._adj.shape[0]
@property
def label_size(self) -> int:
return len(self._label_vocab)
@property
def node_label(self) -> np.ndarray:
return self._node_label
@property
def node_feature_size(self) -> int:
return self._node_feature.shape[1]
@property
def node_feature(self) -> np.ndarray:
return self._node_feature
@property
def edge_size(self) -> int:
return self._adj.nnz // 2
@property
def vocab(self) -> Vocab:
return self._vocab
@property
def node_array(self) -> np.ndarray:
return np.arange(0, self._adj.shape[0])
@property
def edge_array(self) -> np.ndarray:
return np.stack((self._adj.row, self._adj.col), axis=-1)
def read_node_feature(self, feature_dict: dict):
feature_list = [None] * self.node_size
for n, feature in feature_dict.items():
n = self.vocab.stoi[n]
feature_list[n] = feature
feature_array = np.asarray(feature_list, dtype=np.float32)
self._node_feature = feature_array
def read_node_label(self, label_dict: dict):
label_list = [None] * self.node_size
labels = []
for label in label_dict.values():
labels.append(label)
self._label_vocab = Vocab(collections.Counter(labels))
for n, label in label_dict.items():
n = self.vocab.stoi[n]
label = self._label_vocab.stoi[label]
label_list[n] = label
self._node_label = np.asarray(label_list, dtype=np.int32)
def read_edge(self, filename: Path):
if "txt" in filename.suffix:
read_func = read_txt
elif "csv" in filename.suffix:
read_func = read_csv
else:
read_func = read_txt
node_list = list()
for row in read_func(filename):
node_list.append(row[0])
node_list.append(row[1])
self._vocab = Vocab(collections.Counter(node_list))
edge_array = []
for row in read_func(filename):
n1 = self._vocab.stoi[row[0]]
n2 = self._vocab.stoi[row[1]]
edge_array.append([n1, n2])
edge_array.append([n2, n1])
edge_array = np.asarray(edge_array, dtype=np.int32)
self._adj = coo_matrix((np.ones(len(edge_array)), (edge_array[:, 0], edge_array[:, 1])), shape=(len(self._vocab), len(self._vocab)))
self._adj_csr = self._adj.tocsr() + sp.eye(len(self._vocab))
self._adj = self._adj_csr.tocoo()
def get_node_neighbors(self, node: int) -> np.ndarray:
return self._adj_csr[node].indices
def get_node_degree(self, node: int) -> np.ndarray:
return self._adj_csr[node].nnz
def get_nodes_degree_list(self) -> np.ndarray:
return np.asarray([self.get_node_degree(n) for n in self.node_array])
def get_nodes_label(self, nodes: typing.Union[list, np.ndarray, None] = None) -> np.ndarray:
if nodes is None:
return self._node_label
else:
return self._node_label[nodes]
def get_nodes_features(self, nodes: typing.Union[list, np.ndarray, None] = None) -> np.ndarray:
if nodes is None:
return self._node_feature
else:
return self._node_feature[nodes]
class TemporalGraph(StaticGraph):
def __init__(self):
super(TemporalGraph, self).__init__()
self._adj_t = None
self._adj_t_csr = None
self.discrete_g_list = []
self.discrete_adj_list = []
self.discrete_adj_csr_list = []
@property
def adj_t(self) -> sp.coo_matrix:
return self._adj_t
@property
def adj_t_csr(self) -> sp.csr_matrix:
return self._adj_t_csr
@adj_t_csr.setter
def adj_t_csr(self, value: sp.csr_matrix):
self._adj_t_csr = value
self._adj_t = self._adj_t_csr.tocoo()
def discrete(self, slots=10):
nonzero = self.adj_t_csr[self.adj_t_csr.nonzero()]
min_t = np.min(nonzero)
max_t = np.max(nonzero)
slice_ = (max_t - min_t) // slots
adj_t = self.adj_t.toarray()
adj = self.adj.toarray()
for i in range(1, slots + 1):
adj_ = adj.copy()
adj_[np.where(adj_t > min_t + slice_ * i)] = 0
g = StaticGraph()
g.adj = sp.coo_matrix(adj_)
self.discrete_g_list.append(g)
self.discrete_adj_list.append(g.adj)
self.discrete_adj_csr_list.append(g.adj_csr)
def norm(self):
self.adj_t_csr = (self.adj_t_csr - self.adj_t_csr.min()) / (self.adj_t_csr.max() - self.adj_t_csr.min())
def read_edge(self, filename: Path):
if "txt" in filename.suffix:
read_func = read_txt
elif "csv" in filename.suffix:
read_func = read_csv
else:
read_func = read_txt
node_list = list()
for row in read_func(filename):
node_list.append(row[0])
node_list.append(row[1])
self._vocab = Vocab(collections.Counter(node_list))
edge_array = []
for row in read_func(filename):
n1 = self._vocab.stoi[row[0]]
n2 = self._vocab.stoi[row[1]]
t = int(row[2])
edge_array.append([n1, n2, t])
edge_array.append([n2, n1, t])
edge_array = np.asarray(edge_array, dtype=np.int32)
self._adj = coo_matrix((np.ones(len(edge_array)), (edge_array[:, 0], edge_array[:, 1])), shape=(len(self._vocab), len(self._vocab)))
self._adj_t = coo_matrix((edge_array[:, 2], (edge_array[:, 0], edge_array[:, 1])), shape=(len(self._vocab), len(self._vocab)))
self._adj_csr = self._adj.tocsr() + sp.eye(len(self._vocab))
self._adj = self._adj_csr.tocoo()
self._adj_t_csr = self._adj_t.tocsr()