-
Notifications
You must be signed in to change notification settings - Fork 0
/
utils.py
180 lines (155 loc) · 4.72 KB
/
utils.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
import numpy as np
import random
from random import choice
import datetime
from matplotlib.pylab import *
import argparse
import matplotlib.pyplot as plt
from sklearn.decomposition import TruncatedSVD
import networkx as nx
import os
os.chdir('C:/Kaige_Research/Graph Learning/graph_learning_code/')
import pandas as pd
import csv
from sklearn.metrics.pairwise import cosine_similarity, rbf_kernel
from collections import Counter
from scipy.sparse import csgraph
#import seaborn as sns
from scipy.optimize import minimize
from sklearn.preprocessing import MinMaxScaler
from community import community_louvain
from pygsp import graphs, plotting, filters
import pyunlocbox
def sum_squareform(n):
#sum operator that find degree from upper triangle
ncols=int((n-1)*n/2)
I=np.zeros(ncols)
J=np.zeros(ncols)
k=0
for i in list((np.arange(n)+1))[1:]:
I[k:k+(n-i)+1]=(np.arange(n)+1)[i-1:]
k=k+(n-i+1)
k=0
for i in list((np.arange(n)+1))[1:]:
J[k:k+(n-i)+1]=i-1
k=k+n-i+1
I=I.astype(int)
J=J.astype(int)
ys=list(I)+list(J)
xs=list(np.arange(ncols)+1)+list(np.arange(ncols)+1)
s_T=np.zeros((ncols,n))
for i in range(len(ys)):
s_T[(xs[i]-1),(ys[i]-1)]=1
return s_T.T
def vector_form(W,n):
w=W[np.triu_indices(n,1)]
# the triangle-upper
return w
def matrix_form(w, n):
W=np.zeros((n,n))
W[np.triu_indices(n,1)]=w
for i in range(n):
for j in range(n):
W[j,i]=W[i,j]
return W
def scale_0_1(w):
mms=MinMaxScaler()
norm_w=mms.fit_transform(w.reshape(-1,1))
return norm_w.ravel()
def norm_W(W,n):
w=vector_form(W,n)
norm_w=scale_0_1(w)
norm_W=matrix_form(norm_w,n)
return norm_W
def lin_map(x, lims_out, lims_in):
a=lims_in[0]
b=lims_in[1]
c=lims_out[0]
d=lims_out[1]
y=np.zeros(len([x]))
y=((x-a)*(d-c)/(b-a))+c
return y
def filter_graph_to_knn(adj_matrix, node_num, k):
a=adj_matrix.copy()
for i in range(node_num):
rbf_row=a[i,:].ravel()
neighbors=np.argsort(rbf_row)[:node_num-k]
a[i, neighbors]=0
a[neighbors,i]=0
np.fill_diagonal(a,0)
return a
def filter_graph_to_rbf(adj_matrix, node_num, thres=0.5):
adj_matrix[adj_matrix<thres]=0
np.fill_diagonal(adj_matrix,0)
return adj_matrix
def calculate_smoothness(signal, laplacian):
if len(signal)==1:
smooth=np.dot(signal, np.dot(laplacian, signal.T))
else:
smooth=[]
for i in range(len(signal)):
a=signal[i]
s=np.dot(a, np.dot(laplacian, a.T))
smooth.extend([s])
return smooth
def create_networkx_graph(node_num, adj_matrix):
G=nx.Graph()
G.add_nodes_from(list(range(node_num)))
for i in range(node_num):
for j in range(node_num):
if adj_matrix[i,j]!=0:
G.add_edge(i,j,weight=adj_matrix[i,j])
else:
pass
return G
def plot_graph_and_signal(adj_matrix, signal, pos, node_num, error_sigma, title='Graph', path='newpath', show=True):
graph=create_networkx_graph(node_num, adj_matrix)
edge_weight=adj_matrix[np.triu_indices(node_num, 1)]
edge_color=edge_weight[edge_weight>0]
if len(np.unique(edge_color))==1: # unweighted
nodes=nx.draw_networkx_nodes(graph, pos, node_color=signal, node_size=100, cmap=plt.cm.jet)
edges=nx.draw_networkx_edges(graph, pos, width=1.0, alpha=0.3, edge_color='b')
else:# weighted
nodes=nx.draw_networkx_nodes(graph, pos, node_color=signal, node_size=100, cmap=plt.cm.jet)
edges=nx.draw_networkx_edges(graph, pos, width=1.0, alpha=1, edge_color=edge_color, edge_cmap=plt.cm.Blues, vmin=0, vmax=1)
plt.axis('off')
plt.title(title)
plt.savefig(path+title+'.png', dpi=200)
if show==True:
plt.show()
else:
plt.clf()
def generate_graph_from_rbf(adj_matrix):
adj_matrix=np.matrix(adj_matrix)
G=nx.from_numpy_matrix(adj_matrix)
return G
def generate_graph(adj_matrix):
G=nx.Graph()
G.add_nodes_from(list(range(adj_matrix.shape[0])))
for i in range(adj_matrix.shape[0]):
for j in range(adj_matrix.shape[1]):
if adj_matrix[i,j]==0.0:
pass
else:
G.add_edge(i,j, weight=adj_matrix[i,j])
return G
def find_community_best_partition(graph):
parts=community_louvain.best_partition(graph)
values=[parts.get(node) for node in graph.nodes()]
clusters=values
n_clusters=len(np.unique(values))
return clusters, n_clusters
def total_variation_signal_learning(adj, noisy_signal, gamma=3.0):
G=graphs.Graph(adj)
gamma=gamma
d=pyunlocbox.functions.dummy()
r=pyunlocbox.functions.norm_l1()
f=pyunlocbox.functions.norm_l2(w=1, y=noisy_signal, lambda_=gamma)
G.compute_differential_operator()
L=G.D.toarray()
step=0.999/(1+np.linalg.norm(L))
solver=pyunlocbox.solvers.mlfbf(L=L, step=step)
x0=noisy_signal.copy()
prob=pyunlocbox.solvers.solve([d,r,f], solver=solver, x0=x0, rtol=0, maxit=1000)
sol=prob['sol']
return sol