-
Notifications
You must be signed in to change notification settings - Fork 2
/
dataset.py
156 lines (136 loc) · 6.41 KB
/
dataset.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
import torch
import networkx as nx
import torch_geometric.datasets
from torch_geometric.data import Dataset, Data
from torch_geometric.datasets import Planetoid, Amazon, StochasticBlockModelDataset
from torch_geometric.utils import from_networkx, negative_sampling
from torch_scatter import scatter_sum
from utils.utils import index2adjacency, adjacency2index
import urllib.request
import io
import zipfile
import numpy as np
def load_data(configs):
dataset = None
if configs.dataset in ['Cora', 'Citeseer', 'Pubmed', 'Computers', 'Photo']:
dataset = PygDataset(root=configs.root_path, name=configs.dataset)
elif configs.dataset == 'KarateClub':
dataset = KarateClub()
elif configs.dataset == 'FootBall':
dataset = Football()
elif configs.dataset in ['eat', 'bat', 'uat']:
dataset = ATsDataset(root=configs.root_path, name=configs.dataset)
elif configs.dataset == 'SBM':
dataset = SBMDataset(root=configs.root_path)
data = {}
data['feature'] = dataset.feature
data['num_features'] = dataset.num_features
data['edge_index'] = dataset.edge_index
data['degrees'] = dataset.degrees
data['weight'] = dataset.weight
data['num_nodes'] = dataset.num_nodes
data['labels'] = dataset.labels
data['num_classes'] = dataset.num_classes
data['neg_edge_index'] = dataset.neg_edge_index
data['adj'] = dataset.adj
return data
def mask_edges(edge_index, neg_edges, val_prop=0.05, test_prop=0.1):
n = len(edge_index[0])
n_val = int(val_prop * n)
n_test = int(test_prop * n)
edge_val, edge_test, edge_train = edge_index[:, :n_val], edge_index[:, n_val:n_val + n_test], edge_index[:, n_val + n_test:]
val_edges_neg, test_edges_neg = neg_edges[:, :n_val], neg_edges[:, n_val:n_test + n_val]
train_edges_neg = torch.concat([neg_edges, val_edges_neg, test_edges_neg], dim=-1)
return (edge_train, edge_val, edge_test), (train_edges_neg, val_edges_neg, test_edges_neg)
class KarateClub:
def __init__(self):
data = torch_geometric.datasets.KarateClub()
self.feature = data.x
self.num_features = data.x.shape[1]
self.num_nodes = data.x.shape[0]
self.edge_index = data.edge_index
self.weight = torch.ones(self.edge_index.shape[1])
self.degrees = scatter_sum(self.weight, self.edge_index[0])
self.labels = data.y.tolist()
self.num_classes = len(np.unique(self.labels))
self.neg_edge_index = negative_sampling(data.edge_index)
self.adj = index2adjacency(self.num_nodes, self.edge_index, self.weight, is_sparse=True)
class Football:
def __init__(self):
url = "http://www-personal.umich.edu/~mejn/netdata/football.zip"
sock = urllib.request.urlopen(url) # open URL
s = io.BytesIO(sock.read()) # read into BytesIO "file"
sock.close()
zf = zipfile.ZipFile(s) # zipfile object
txt = zf.read("football.txt").decode() # read info file
gml = zf.read("football.gml").decode() # read gml data
# throw away bogus first line with # from mejn files
gml = gml.split("\n")[1:]
graph = nx.parse_gml(gml) # parse gml data
data = from_networkx(graph)
self.feature = torch.eye(data.num_nodes)
self.num_features = data.num_nodes
self.num_nodes = data.num_nodes
self.edge_index = data.edge_index
self.weight = torch.ones(self.edge_index.shape[1])
self.degrees = scatter_sum(self.weight, self.edge_index[0])
self.labels = data.value.tolist()
self.num_classes = len(np.unique(self.labels))
self.neg_edge_index = negative_sampling(data.edge_index)
self.adj = index2adjacency(self.num_nodes, self.edge_index, self.weight, is_sparse=True)
class PygDataset:
def __init__(self, root, name='Cora'):
if name in ['Cora', 'Citeseer', 'Pubmed']:
dataset = Planetoid(root, name)
else:
dataset = Amazon(root, name)
data = dataset.data
self.num_nodes = data.x.shape[0]
self.feature = data.x
self.num_features = data.x.shape[1]
self.edge_index = data.edge_index
self.weight = torch.ones(self.edge_index.shape[1])
self.degrees = scatter_sum(self.weight, self.edge_index[0])
self.labels = data.y.tolist()
self.num_classes = len(np.unique(self.labels))
self.neg_edge_index = negative_sampling(data.edge_index)
self.adj = index2adjacency(self.num_nodes, self.edge_index, self.weight, is_sparse=True)
class ATsDataset:
def __init__(self, root, name='eat'):
adj = np.load(f'{root}/{name}/{name}_adj.npy')
feat = np.load(f'{root}/{name}/{name}_feat.npy')
label = np.load(f'{root}/{name}/{name}_label.npy')
self.num_nodes = feat.shape[0]
self.feature = torch.tensor(feat).float()
self.num_features = feat.shape[1]
self.edge_index = adjacency2index(torch.tensor(adj))
self.weight = torch.ones(self.edge_index.shape[1])
self.degrees = scatter_sum(self.weight, self.edge_index[0])
self.labels = list(label)
self.num_classes = len(np.unique(self.labels))
self.neg_edge_index = negative_sampling(self.edge_index)
self.adj = index2adjacency(self.num_nodes, self.edge_index, self.weight, is_sparse=True)
class SBMDataset:
def __init__(self, root, num_classes=5, num_nodes=200, p_in=0.6, p_out=0.03):
num_classes = num_classes
p = torch.zeros(num_classes, num_classes)
for i in range(num_classes):
for j in range(num_classes):
if i == j:
p[i, j] = p_in
else:
p[i, j] = p_out
data = StochasticBlockModelDataset(root,
[num_nodes / num_classes] * num_classes,
p, num_nodes=num_nodes)[0]
data.x = torch.eye(data.num_nodes)
self.num_nodes = data.x.shape[0]
self.feature = data.x
self.num_features = data.x.shape[1]
self.edge_index = data.edge_index
self.weight = torch.ones(self.edge_index.shape[1])
self.degrees = scatter_sum(self.weight, self.edge_index[0])
self.labels = data.y.tolist()
self.num_classes = len(np.unique(self.labels))
self.neg_edge_index = negative_sampling(data.edge_index)
self.adj = index2adjacency(self.num_nodes, self.edge_index, self.weight, is_sparse=True)