RAG-GFM/aug.py at main · RingBDStack/RAG-GFM · GitHub

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
import torch
import copy
import random
import pdb
import scipy.sparse as sp
import numpy as np

def main():
    pass


def \
        aug_random_mask(input_feature, drop_percent=0.2):

    node_num = input_feature.shape[1]
    mask_num = int(node_num * drop_percent)
    node_idx = [i for i in range(node_num)]
    mask_idx = random.sample(node_idx, mask_num)
    aug_feature = copy.deepcopy(input_feature)
    zeros = torch.zeros_like(aug_feature[0][0])
    for j in mask_idx:
        aug_feature[0][j] = zeros
    return aug_feature


def aug_random_edge(input_adj, drop_percent=0.2):

    percent = drop_percent / 2
    row_idx, col_idx = input_adj.nonzero()

    index_list = []
    for i in range(len(row_idx)):
        index_list.append((row_idx[i], col_idx[i]))

    single_index_list = []
    for i in list(index_list):
        single_index_list.append(i)
        index_list.remove((i[1], i[0]))


    edge_num = int(len(row_idx) / 2)
    add_drop_num = int(edge_num * percent / 2)
    aug_adj = copy.deepcopy(input_adj.todense().tolist())

    edge_idx = [i for i in range(edge_num)]
    drop_idx = random.sample(edge_idx, add_drop_num)


    for i in drop_idx:
        aug_adj[single_index_list[i][0]][single_index_list[i][1]] = 0
        aug_adj[single_index_list[i][1]][single_index_list[i][0]] = 0

    '''
    above finish drop edges
    '''
    node_num = input_adj.shape[0]
    l = [(i, j) for i in range(node_num) for j in range(i)]
    add_list = random.sample(l, add_drop_num)

    for i in add_list:

        aug_adj[i[0]][i[1]] = 1
        aug_adj[i[1]][i[0]] = 1

    aug_adj = np.matrix(aug_adj)
    aug_adj = sp.csr_matrix(aug_adj)
    return aug_adj


def aug_drop_node(input_fea, input_adj, drop_percent=0.2):

    input_adj = torch.tensor(input_adj.todense().tolist())
    input_fea = input_fea.squeeze(0)

    node_num = input_fea.shape[0]
    drop_num = int(node_num * drop_percent)
    all_node_list = [i for i in range(node_num)]

    drop_node_list = sorted(random.sample(all_node_list, drop_num))

    aug_input_fea = delete_row_col(input_fea, drop_node_list, only_row=True)
    aug_input_adj = delete_row_col(input_adj, drop_node_list)

    aug_input_fea = aug_input_fea.unsqueeze(0)
    aug_input_adj = sp.csr_matrix(np.matrix(aug_input_adj))

    return aug_input_fea, aug_input_adj


def aug_subgraph(input_fea, input_adj, drop_percent=0.2):

    input_adj = torch.tensor(input_adj.todense().tolist())
    input_fea = input_fea.squeeze(0)
    node_num = input_fea.shape[0]

    all_node_list = [i for i in range(node_num)]
    s_node_num = int(node_num * (1 - drop_percent))
    center_node_id = random.randint(0, node_num - 1)
    sub_node_id_list = [center_node_id]
    all_neighbor_list = []

    for i in range(s_node_num - 1):

        all_neighbor_list += torch.nonzero(input_adj[sub_node_id_list[i]], as_tuple=False).squeeze(1).tolist()

        all_neighbor_list = list(set(all_neighbor_list))
        new_neighbor_list = [n for n in all_neighbor_list if not n in sub_node_id_list]
        if len(new_neighbor_list) != 0:
            new_node = random.sample(new_neighbor_list, 1)[0]
            sub_node_id_list.append(new_node)
        else:
            break


    drop_node_list = sorted([i for i in all_node_list if not i in sub_node_id_list])

    aug_input_fea = delete_row_col(input_fea, drop_node_list, only_row=True)
    aug_input_adj = delete_row_col(input_adj, drop_node_list)

    aug_input_fea = aug_input_fea.unsqueeze(0)
    aug_input_adj = sp.csr_matrix(np.matrix(aug_input_adj))

    return aug_input_fea, aug_input_adj


def delete_row_col(input_matrix, drop_list, only_row=False):

    remain_list = [i for i in range(input_matrix.shape[0]) if i not in drop_list]
    out = input_matrix[remain_list, :]
    if only_row:
        return out
    out = out[:, remain_list]

    return out


if __name__ == "__main__":
    main()