Maui/load_datasets.py at main · iHeartGraph/Maui · 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
160
import os
import torch
import scipy.sparse as sp
import numpy as np

from sklearn.metrics import precision_score, recall_score
from torch_geometric.datasets import Planetoid, Amazon, GitHub, FacebookPagePage, LastFMAsia, \
    DeezerEurope, WikiCS, Flickr, Twitch, Coauthor
from torch_geometric.utils import to_dense_adj, sort_edge_index, to_scipy_sparse_matrix, from_scipy_sparse_matrix
from torch_geometric.data import Data

rng = np.random.default_rng(0)


def get_dataset(dataset_dir: str, dataset_name: str, epsilon: float = None):
    """load dataset"""
    if dataset_name.lower() in ['cora', 'citeseer', 'pubmed']:
        dataset = Planetoid(root=dataset_dir, name=dataset_name)
    elif dataset_name.lower() in ['computers', 'photo']:
        dataset = Amazon(root=dataset_dir, name=dataset_name)
    elif dataset_name.lower() == 'github':
        dataset_dir = os.path.join(dataset_dir, dataset_name)
        dataset = GitHub(root=dataset_dir)
    elif dataset_name.lower() == 'facebook':
        dataset_dir = os.path.join(dataset_dir, dataset_name)
        dataset = FacebookPagePage(root=dataset_dir)
    elif dataset_name.lower() == 'lastfmasia':
        dataset_dir = os.path.join(dataset_dir, dataset_name)
        dataset = LastFMAsia(root=dataset_dir)
    elif dataset_name.lower() == 'deezereurope':
        dataset_dir = os.path.join(dataset_dir, dataset_name)
        dataset = DeezerEurope(root=dataset_dir)
    elif dataset_name.lower() == 'wikics':
        dataset_dir = os.path.join(dataset_dir, dataset_name)
        dataset = WikiCS(root=dataset_dir, is_undirected=True)
    elif dataset_name.lower() == 'flicker':
        dataset_dir = os.path.join(dataset_dir, dataset_name)
        dataset = Flickr(root=dataset_dir)
    elif dataset_name.lower() in ['de', 'en', 'es', 'fr', 'pt', 'ru']:
        dataset = Twitch(root=dataset_dir, name=dataset_name.upper())
    elif dataset_name.lower() in ['cs', 'physics']:
        dataset = Coauthor(root=dataset_dir, name=dataset_name)
    else:
        raise NotImplementedError

    '''split data'''
    data = dataset[0]
    num_train = int(data.num_nodes * 0.6)
    num_val = int(data.num_nodes * 0.2)
    data.train_mask = torch.zeros(data.num_nodes, dtype=torch.bool)
    data.val_mask = torch.zeros(data.num_nodes, dtype=torch.bool)
    data.test_mask = torch.zeros(data.num_nodes, dtype=torch.bool)
    perm_index = torch.randperm(data.num_nodes, generator=torch.random.manual_seed(0))
    data.train_mask[perm_index[:num_train]] = True
    data.val_mask[perm_index[num_train:num_train + num_val]] = True
    data.test_mask[perm_index[num_train + num_val:]] = True

    data.true_num_edges = data.num_edges

    if epsilon is None:
        dataset.data = data
        return dataset

    '''differential privacy'''
    adj = to_scipy_sparse_matrix(data.edge_index, num_nodes=data.num_nodes).tocoo()
    adj_noisy = perturb_adj_continuous(adj, epsilon)
    edge_index = from_scipy_sparse_matrix(adj_noisy)[0]

    data.edge_index = edge_index
    dataset.data = data

    return dataset


def construct_sparse_mat(indice, N):
    cur_row = -1
    new_indices = []
    new_indptr = []

    for i, j in indice:
        if i >= j:
            continue

        while i > cur_row:
            new_indptr.append(len(new_indices))
            cur_row += 1

        new_indices.append(j)

    while N > cur_row:
        new_indptr.append(len(new_indices))
        cur_row += 1

    data = np.ones(len(new_indices), dtype=np.int64)
    indices = np.asarray(new_indices, dtype=np.int64)
    indptr = np.asarray(new_indptr, dtype=np.int64)

    mat = sp.csr_matrix((data, indices, indptr), (N, N))

    return mat + mat.T


def perturb_adj_continuous(adj, epsilon):
    n_nodes = adj.shape[0]
    n_edges = len(adj.data) // 2

    N = n_nodes

    A = sp.tril(adj, k=-1)

    eps_1 = epsilon * 0.01
    eps_2 = epsilon - eps_1
    noise = get_noise(noise_type='laplace', size=(N, N),
                      eps=eps_2, delta=1e-5, sensitivity=1)
    noise *= np.tri(*noise.shape, k=-1, dtype=np.bool)
    A += noise
    n_edges_keep = n_edges + int(
        get_noise(noise_type='laplace', size=1,
                  eps=eps_1, delta=1e-5, sensitivity=1)[0])
    a_r = A.A.ravel()

    n_splits = 50
    len_h = len(a_r) // n_splits
    ind_list = []
    for i in range(n_splits - 1):
        ind = np.argpartition(a_r[len_h * i:len_h * (i + 1)], -n_edges_keep)[-n_edges_keep:]
        ind_list.append(ind + len_h * i)

    ind = np.argpartition(a_r[len_h * (n_splits - 1):], -n_edges_keep)[-n_edges_keep:]
    ind_list.append(ind + len_h * (n_splits - 1))

    ind_subset = np.hstack(ind_list)
    a_subset = a_r[ind_subset]
    ind = np.argpartition(a_subset, -n_edges_keep)[-n_edges_keep:]

    row_idx = []
    col_idx = []
    for idx in ind:
        idx = ind_subset[idx]
        row_idx.append(idx // N)
        col_idx.append(idx % N)
        assert (col_idx < row_idx)
    data_idx = np.ones(n_edges_keep, dtype=np.int32)

    mat = sp.csr_matrix((data_idx, (row_idx, col_idx)), shape=(N, N))
    return mat + mat.T


def get_noise(noise_type, size, eps=10, delta=1e-5, sensitivity=2):

    if noise_type == 'laplace':
        noise = rng.laplace(0, sensitivity/eps, size)
    elif noise_type == 'gaussian':
        c = np.sqrt(2*np.log(1.25/delta))
        stddev = c * sensitivity / eps
        noise = rng.normal(0, stddev, size)
    else:
        raise NotImplementedError('noise {} not implemented!'.format(noise_type))

    return noise