-
Notifications
You must be signed in to change notification settings - Fork 1
/
data_split.py
152 lines (118 loc) · 4.99 KB
/
data_split.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
import torch
from ogb.nodeproppred import PygNodePropPredDataset
from torch_geometric.datasets import CoraFull, Reddit2, Coauthor, Planetoid, Amazon, DBLP
import random
import numpy as np
import scipy.io as sio
from sklearn import preprocessing
# class_split = {"train": 0.6,"test": 0.4}
class_split = {
"CoraFull": {"train": 40, 'dev': 15, 'test': 15}, # Sufficient number of base classes
"ogbn-arxiv": {"train": 20, 'dev': 10, 'test': 10},
"Coauthor-CS": {"train": 5, 'dev': 5, 'test': 5},
"Amazon-Computer": {"train": 4, 'dev': 3, 'test': 3},
"Cora": {"train": 3, 'dev': 2, 'test': 2},
"CiteSeer": {"train": 2, 'dev': 2, 'test': 2},
"Reddit": {"train": 21, 'dev': 10, 'test': 10},
'dblp':{"train": 77, 'dev': 30, 'test': 30},
}
class dblp_data():
def __init__(self):
self.x=None
self.edge_index=None
self.num_nodes=None
self.y=None
self.num_edges=None
self.num_features=7202
class dblp_dataset():
def __init__(self,data,num_classes):
self.data=data
self.num_classes=num_classes
def load_DBLP(root=None, dataset_source='dblp'):
dataset=dblp_data()
n1s = []
n2s = []
for line in open("./few_shot_data/{}_network".format(dataset_source)):
n1, n2 = line.strip().split('\t')
if int(n1)>int(n2):
n1s.append(int(n1))
n2s.append(int(n2))
num_nodes = max(max(n1s), max(n2s)) + 1
print('nodes num', num_nodes)
data_train = sio.loadmat("./few_shot_data/{}_train.mat".format(dataset_source))
data_test = sio.loadmat("./few_shot_data/{}_test.mat".format(dataset_source))
labels = np.zeros((num_nodes, 1))
labels[data_train['Index']] = data_train["Label"]
labels[data_test['Index']] = data_test["Label"]
features = np.zeros((num_nodes, data_train["Attributes"].shape[1]))
features[data_train['Index']] = data_train["Attributes"].toarray()
features[data_test['Index']] = data_test["Attributes"].toarray()
lb = preprocessing.LabelBinarizer()
labels = lb.fit_transform(labels)
features = torch.FloatTensor(features)
labels = torch.LongTensor(np.where(labels)[1])
dataset.edge_index=torch.tensor([n1s,n2s])
dataset.y=labels
dataset.x=features
dataset.num_nodes=num_nodes
dataset.num_edges=dataset.edge_index.shape[1]
return dblp_dataset(dataset, num_classes=80+27+30)
def split(dataset_name):
if dataset_name == 'Cora':
dataset = Planetoid(root='~/dataset/' + dataset_name, name="Cora")
num_nodes = dataset.data.num_nodes
elif dataset_name == 'CiteSeer':
dataset = Planetoid(root='~/dataset/' + dataset_name, name="CiteSeer")
num_nodes = dataset.data.num_nodes
elif dataset_name == 'Amazon-Computer':
dataset = Amazon(root='~/dataset/' + dataset_name, name="Computers")
num_nodes = dataset.data.num_nodes
elif dataset_name == 'Coauthor-CS':
dataset = Coauthor(root='~/dataset/' + dataset_name, name="CS")
num_nodes = dataset.data.num_nodes
elif dataset_name == 'CoraFull':
dataset = CoraFull(root='./dataset/' + dataset_name)
num_nodes = dataset.data.num_nodes
elif dataset_name == 'Reddit':
dataset = Reddit2(root='./dataset/' + dataset_name)
num_nodes = dataset.data.num_nodes
elif dataset_name == 'ogbn-arxiv':
dataset = PygNodePropPredDataset(name = dataset_name, root='./dataset/' + dataset_name)
num_nodes = dataset.data.num_nodes
elif dataset_name == 'dblp':
dataset = load_DBLP(root='./few_shot_data/')
num_nodes=dataset.data.num_nodes
else:
print("Dataset not support!")
exit(0)
data = dataset.data
class_list = [i for i in range(dataset.num_classes)]
print("********" * 10)
train_num = class_split[dataset_name]["train"]
dev_num = class_split[dataset_name]["dev"]
test_num = class_split[dataset_name]["test"]
random.shuffle(class_list)
train_class = class_list[: train_num]
dev_class = class_list[train_num : train_num + dev_num]
test_class = class_list[train_num + dev_num :]
print("train_num: {}; dev_num: {}; test_num: {}".format(train_num, dev_num, test_num))
id_by_class = {}
for i in class_list:
id_by_class[i] = []
for id, cla in enumerate(torch.squeeze(data.y).tolist()):
id_by_class[cla].append(id)
train_idx = []
for cla in train_class:
train_idx.extend(id_by_class[cla])
degree_inv = num_nodes / (dataset.data.num_edges * 2)
return data, np.array(train_idx), id_by_class, train_class, dev_class, test_class, degree_inv
def test_task_generator(id_by_class, class_list, n_way, k_shot, m_query):
# sample class indices
class_selected = random.sample(class_list, n_way)
id_support = []
id_query = []
for cla in class_selected:
temp = random.sample(id_by_class[cla], k_shot + m_query)
id_support.extend(temp[:k_shot])
id_query.extend(temp[k_shot:])
return np.array(id_support), np.array(id_query), class_selected