-
Notifications
You must be signed in to change notification settings - Fork 0
/
datasets.py
212 lines (167 loc) · 7.6 KB
/
datasets.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
import os.path as osp
from typing import Callable, List, Optional
import numpy as np
from scipy import io as sio
import torch
from torch_geometric.data import (HeteroData, InMemoryDataset)
from torch_geometric.datasets import DBLP, IMDB
from torch_geometric import transforms as T
class ACM(InMemoryDataset):
def __init__(self, root: str, transform: Optional[Callable] = None,
pre_transform: Optional[Callable] = None):
super().__init__(root, transform, pre_transform)
self.data, self.slices = torch.load(self.processed_paths[0])
@property
def raw_file_names(self) -> List[str]:
return ['ACM.mat']
@property
def processed_file_names(self) -> str:
return 'data.pt'
def process(self):
data = HeteroData()
raw_data = sio.loadmat(osp.join(self.raw_dir, 'ACM.mat'))
p_vs_l = raw_data['PvsL']
p_vs_a = raw_data['PvsA']
p_vs_t = raw_data['PvsT']
p_vs_p = raw_data['PvsP']
p_vs_c = raw_data['PvsC']
conf_ids = [0, 1, 9, 10, 13]
label_ids = [0, 1, 2, 2, 1]
p_vs_c_filter = p_vs_c[:, conf_ids]
p_selected = (p_vs_c_filter.sum(1) != 0).A1.nonzero()[0]
p_vs_c = p_vs_c[p_selected]
p_vs_p = p_vs_p[p_selected].T[p_selected]
a_selected = (p_vs_a[p_selected].sum(0) != 0).A1.nonzero()[0]
p_vs_a = p_vs_a[p_selected].T[a_selected].T
l_selected = (p_vs_l[p_selected].sum(0) != 0).A1.nonzero()[0]
p_vs_l = p_vs_l[p_selected].T[l_selected].T
t_selected = (p_vs_t[p_selected].sum(0) != 0).A1.nonzero()[0]
p_vs_t = p_vs_t[p_selected].T[t_selected].T
pc_p, pc_c = p_vs_c.nonzero()
labels = np.zeros(len(p_selected), dtype=np.int64)
for conf_id, label_id in zip(conf_ids, label_ids):
labels[pc_p[pc_c == conf_id]] = label_id
labels = torch.LongTensor(labels)
data['paper'].x = torch.FloatTensor(p_vs_t.toarray())
data['paper'].y = torch.LongTensor(labels)
data['paper', 'author'].edge_index = torch.tensor(p_vs_a.nonzero(), dtype=torch.long)
data['author', 'paper'].edge_index = torch.tensor(p_vs_a.transpose().nonzero(), dtype=torch.long)
data['paper', 'subject'].edge_index = torch.tensor(p_vs_l.nonzero(), dtype=torch.long)
data['subject', 'paper'].edge_index = torch.tensor(p_vs_l.transpose().nonzero(), dtype=torch.long)
if self.pre_transform is not None:
data = self.pre_transform(data)
torch.save(self.collate([data]), self.processed_paths[0])
def __repr__(self) -> str:
return f'{self.__class__.__name__}()'
class AMiner(InMemoryDataset):
def __init__(self, root: str, transform: Optional[Callable] = None,
pre_transform: Optional[Callable] = None):
super().__init__(root, transform, pre_transform)
self.data, self.slices = torch.load(self.processed_paths[0])
@property
def raw_file_names(self) -> List[str]:
return ['labels.npy', 'pr.txt', 'pa.txt',
'feature_0.npy', 'feature_1.npy', 'feature_2.npy']
@property
def processed_file_names(self) -> str:
return 'data.pt'
def process(self):
data = HeteroData()
node_types = ['paper', 'author', 'reference']
for i, node_type in enumerate(node_types):
x = np.load(osp.join(self.raw_dir, f'features_{i}.npy'))
data[node_type].x = torch.from_numpy(x).to(torch.float)
labels = np.load(osp.join(self.raw_dir, 'labels.npy')).astype('int32')
data['paper'].y = torch.from_numpy(labels)
pa = np.loadtxt(osp.join(self.raw_dir, 'pa.txt'))
pa = torch.from_numpy(pa).t()
pr = np.loadtxt(osp.join(self.raw_dir, 'pr.txt'))
pr = torch.from_numpy(pr).t()
data['paper', 'reference'].edge_index = pr[[0, 1]].long()
data['reference', 'paper'].edge_index = pr[[1, 0]].long()
data['paper', 'author'].edge_index = pa[[0, 1]].long()
data['author', 'paper'].edge_index = pa[[1, 0]].long()
if self.pre_transform is not None:
data = self.pre_transform(data)
torch.save(self.collate([data]), self.processed_paths[0])
def __repr__(self) -> str:
return f'{self.__class__.__name__}()'
class FreeBase(InMemoryDataset):
def __init__(self, root: str, transform: Optional[Callable] = None,
pre_transform: Optional[Callable] = None):
super().__init__(root, transform, pre_transform)
self.data, self.slices = torch.load(self.processed_paths[0])
@property
def raw_file_names(self) -> List[str]:
return ['labels.npy', 'ma.txt', 'md.txt', 'mw.txt', 'feature_0.npy',
'feature_1.npy', 'feature_2.npy', 'feature_3.npy']
@property
def processed_file_names(self) -> str:
return 'data.pt'
def process(self):
data = HeteroData()
node_types = ['movie', 'actor', 'director', 'writer']
for i, node_type in enumerate(node_types):
x = np.load(osp.join(self.raw_dir, f'features_{i}.npy'))
data[node_type].x = torch.from_numpy(x).to(torch.float)
labels = np.load(osp.join(self.raw_dir, 'labels.npy')).astype('int32')
data['movie'].y = torch.from_numpy(labels)
ma = np.loadtxt(osp.join(self.raw_dir, 'ma.txt'))
ma = torch.from_numpy(ma).t()
md = np.loadtxt(osp.join(self.raw_dir, 'md.txt'))
md = torch.from_numpy(md).t()
mw = np.loadtxt(osp.join(self.raw_dir, 'mw.txt'))
mw = torch.from_numpy(mw).t()
data['movie', 'actor'].edge_index = ma[[0, 1]].long()
data['actor', 'movie'].edge_index = ma[[1, 0]].long()
data['movie', 'director'].edge_index = md[[0, 1]].long()
data['director', 'movie'].edge_index = md[[1, 0]].long()
data['movie', 'writer'].edge_index = mw[[0, 1]].long()
data['writer', 'movie'].edge_index = mw[[1, 0]].long()
if self.pre_transform is not None:
data = self.pre_transform(data)
torch.save(self.collate([data]), self.processed_paths[0])
def __repr__(self) -> str:
return f'{self.__class__.__name__}()'
def get_dataset(dataset_name):
path = osp.join('data', dataset_name)
if dataset_name == 'dblp':
dataset = DBLP(path)
metapaths = [
[('author', 'paper'), ('paper', 'author')],
[('author', 'paper'), ('paper', 'term'), ('term', 'paper'), ('paper', 'author')],
[('author', 'paper'), ('paper', 'conference'), ('conference', 'paper'), ('paper', 'author')]
]
target = 'author'
elif dataset_name == 'imdb':
dataset = IMDB(path)
metapaths = [
[('movie', 'director'), ('director', 'movie')],
[('movie', 'actor'), ('actor', 'movie')]
]
target = 'movie'
elif dataset_name == 'acm':
dataset = ACM(path)
metapaths = [
[('paper', 'author'), ('author', 'paper')],
[('paper', 'subject'), ('subject', 'paper')]
]
target = 'paper'
elif dataset_name == 'aminer':
dataset = AMiner(path)
metapaths = [
[('paper', 'reference'), ('reference', 'paper')],
[('paper', 'author'), ('author', 'paper')]
]
target = 'paper'
elif dataset_name == 'freebase':
dataset = FreeBase(path)
metapaths = [
[('movie', 'actor'), ('actor', 'movie')],
[('movie', 'director'), ('director', 'movie')],
[('movie', 'writer'), ('writer', 'movie')]
]
target = 'movie'
else:
raise TypeError('Unsupported dataset!')
return dataset, metapaths, target