This repository has been archived by the owner on Sep 20, 2022. It is now read-only.
-
Notifications
You must be signed in to change notification settings - Fork 2
/
freq_resp.py
205 lines (173 loc) · 7.83 KB
/
freq_resp.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
import os
import numpy as np
import scipy.sparse as sp
from scipy.linalg import eigvalsh
def norm_feat(feature):
feature = feature.astype(dtype=np.float64)
if sp.issparse(feature):
row_sum = feature.sum(axis=1).A1
row_sum_inv = np.power(row_sum, -1)
row_sum_inv[np.isinf(row_sum_inv)] = 0.
deg_inv = sp.diags(row_sum_inv, format='csc')
norm_feature = deg_inv.dot(feature)
else:
row_sum_inv = np.power(np.sum(feature, axis=1), -1)
row_sum_inv[np.isinf(row_sum_inv)] = 0.
deg_inv = np.diag(row_sum_inv)
norm_feature = deg_inv.dot(feature)
norm_feature = np.array(norm_feature, dtype=np.float64)
return norm_feature
def load_webkb_data(dataset_name):
dataset_path = './data'
dataset_path = os.path.join(dataset_path, dataset_name)
feature = sp.load_npz(os.path.join(dataset_path, 'features.npz'))
feature = feature.tocsc()
n_feat = feature.shape[1]
feature = norm_feat(feature)
feature = feature.astype(dtype=np.float32)
adj = sp.load_npz(os.path.join(dataset_path, 'adj.npz'))
adj = adj.tocsc()
label = np.genfromtxt(os.path.join(dataset_path, 'labels.csv'))
n_class = 5
idx_train = np.genfromtxt(os.path.join(dataset_path, 'idx_train.csv'))
idx_valid = np.genfromtxt(os.path.join(dataset_path, 'idx_valid.csv'))
idx_test = np.genfromtxt(os.path.join(dataset_path, 'idx_test.csv'))
return feature, adj, label, idx_train, idx_valid, idx_test, n_feat, n_class
def calc_mag_gso(dir_adj, gso_type, q):
if sp.issparse(dir_adj):
id = sp.identity(dir_adj.shape[0], format='csc')
# Symmetrizing an adjacency matrix
adj = dir_adj + dir_adj.T.multiply(dir_adj.T > dir_adj) - dir_adj.multiply(dir_adj.T > dir_adj)
#adj = 0.5 * (dir_adj + dir_adj.transpose())
if q != 0:
dir = dir_adj.transpose() - dir_adj
trs = np.exp(1j * 2 * np.pi * q * dir.toarray())
trs = sp.csc_matrix(trs)
else:
trs = id # Fake
if gso_type == 'sym_renorm_mag_adj' or gso_type == 'rw_renorm_mag_adj' \
or gso_type == 'neg_sym_renorm_mag_adj' or gso_type == 'neg_rw_renorm_mag_adj' \
or gso_type == 'sym_renorm_mag_lap' or gso_type == 'rw_renorm_mag_lap':
adj = adj + id
if gso_type == 'sym_norm_mag_adj' or gso_type == 'sym_renorm_mag_adj' \
or gso_type == 'neg_sym_norm_mag_adj' or gso_type == 'neg_sym_renorm_mag_adj' \
or gso_type == 'sym_norm_mag_lap' or gso_type == 'sym_renorm_mag_lap':
row_sum = adj.sum(axis=1).A1
row_sum_inv_sqrt = np.power(row_sum, -0.5)
row_sum_inv_sqrt[np.isinf(row_sum_inv_sqrt)] = 0.
deg_inv_sqrt = sp.diags(row_sum_inv_sqrt, format='csc')
# A_{sym} = D^{-0.5} * A * D^{-0.5}
sym_norm_adj = deg_inv_sqrt.dot(adj).dot(deg_inv_sqrt)
if q == 0:
sym_norm_mag_adj = sym_norm_adj
elif q == 0.5:
sym_norm_mag_adj = sym_norm_adj.multiply(trs.real)
else:
sym_norm_mag_adj = sym_norm_adj.multiply(trs)
if gso_type == 'neg_sym_norm_mag_adj' or gso_type == 'neg_sym_renorm_mag_adj':
gso = -1 * sym_norm_mag_adj
elif gso_type == 'sym_norm_mag_lap' or gso_type == 'sym_renorm_mag_lap':
sym_norm_mag_lap = id - sym_norm_mag_adj
gso = sym_norm_mag_lap
else:
gso = sym_norm_mag_adj
elif gso_type == 'rw_norm_mag_adj' or gso_type == 'rw_renorm_mag_adj' \
or gso_type == 'neg_rw_norm_mag_adj' or gso_type == 'neg_rw_renorm_mag_adj' \
or gso_type == 'rw_norm_mag_lap' or gso_type == 'rw_renorm_mag_lap':
row_sum = adj.sum(axis=1).A1
row_sum_inv = np.power(row_sum, -1)
row_sum_inv[np.isinf(row_sum_inv)] = 0.
deg_inv = sp.diags(row_sum_inv, format='csc')
# A_{rw} = D^{-1} * A
rw_norm_adj = deg_inv.dot(adj)
if q == 0:
rw_norm_mag_adj = rw_norm_adj
elif q == 0.5:
rw_norm_mag_adj = rw_norm_adj.multiply(trs.real)
else:
rw_norm_mag_adj = rw_norm_adj.multiply(trs)
if gso_type == 'neg_rw_norm_mag_adj' or gso_type == 'neg_rw_renorm_mag_adj':
gso = -1 * rw_norm_mag_adj
elif gso_type == 'rw_norm_mag_lap' or gso_type == 'rw_renorm_mag_lap':
rw_norm_mag_lap = id - rw_norm_mag_adj
gso = rw_norm_mag_lap
else:
gso = rw_norm_mag_adj
else:
raise ValueError(f'{gso_type} is not defined.')
else:
id = np.identity(dir_adj.shape[0])
# Symmetrizing an adjacency matrix
adj = np.maximum(dir_adj, dir_adj.T)
#adj = 0.5 * (dir_adj + dir_adj.T)
if q != 0:
dir = dir_adj.T - dir_adj
trs = np.exp(1j * 2 * np.pi * q * dir)
else:
trs = id # Fake
if gso_type == 'sym_renorm_mag_adj' or gso_type == 'rw_renorm_mag_adj' \
or gso_type == 'sym_renorm_mag_lap' or gso_type == 'rw_renorm_mag_lap':
adj = adj + id
if gso_type == 'sym_norm_mag_adj' or gso_type == 'sym_renorm_mag_adj' \
or gso_type == 'sym_norm_mag_lap' or gso_type == 'sym_renorm_mag_lap':
row_sum = np.sum(adj, axis=1)
row_sum_inv_sqrt = np.power(row_sum, -0.5)
row_sum_inv_sqrt[np.isinf(row_sum_inv_sqrt)] = 0.
deg_inv_sqrt = np.diag(row_sum_inv_sqrt)
# A_{sym} = D^{-0.5} * A * D^{-0.5}
sym_norm_adj = deg_inv_sqrt.dot(adj).dot(deg_inv_sqrt)
if q == 0:
sym_norm_mag_adj = sym_norm_adj
elif q == 0.5:
sym_norm_mag_adj = np.multiply(sym_norm_adj, trs.real)
else:
sym_norm_mag_adj = np.multiply(sym_norm_adj, trs)
if gso_type == 'sym_norm_mag_lap' or gso_type == 'sym_renorm_mag_lap':
sym_norm_mag_lap = id - sym_norm_mag_adj
gso = sym_norm_mag_lap
else:
gso = sym_norm_mag_adj
elif gso_type == 'rw_norm_mag_adj' or gso_type == 'rw_renorm_mag_adj' \
or gso_type == 'rw_norm_mag_lap' or gso_type == 'rw_renorm_mag_lap':
row_sum = np.sum(adj, axis=1).A1
row_sum_inv = np.power(row_sum, -1)
row_sum_inv[np.isinf(row_sum_inv)] = 0.
deg_inv = np.diag(row_sum_inv)
# A_{rw} = D^{-1} * A
rw_norm_adj = deg_inv.dot(adj)
if q == 0:
rw_norm_mag_adj = rw_norm_adj
elif q == 0.5:
rw_norm_mag_adj = np.multiply(rw_norm_adj, trs.real)
else:
rw_norm_mag_adj = np.multiply(rw_norm_adj, trs)
if gso_type == 'rw_norm_mag_lap' or gso_type == 'rw_renorm_mag_lap':
rw_norm_mag_lap = id - rw_norm_mag_adj
gso = rw_norm_mag_lap
else:
gso = rw_norm_mag_adj
else:
raise ValueError(f'{gso_type} is not defined.')
return gso
dataset = 'wisconsin'
gso_type = 'sym_renorm_mag_lap'
q = 1/6
feature, adj, label, idx_train, idx_val, idx_test, n_feat, n_class = load_webkb_data(dataset)
gso = calc_mag_gso(adj, gso_type, q)
if sp.issparse(gso):
gso = gso.toarray()
if q == 0 or q == 0.5:
gso = gso.astype(np.float64)
else:
gso = gso.astype(np.complex128)
eigval = eigvalsh(a=gso).real
eigval = np.sort(eigval)
else:
if q == 0 or q == 0.5:
gso = gso.astype(np.float64)
else:
gso = gso.astype(np.complex128)
eigval = eigvalsh(a=gso).real
eigval = np.sort(eigval)
csv_name = dataset + '_sym_renorm_mag_lap_' + str(q) + '_ev.csv'
np.savetxt(fname=csv_name, X=eigval)