-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcd_knn_gl_MAB_test.py
190 lines (159 loc) · 7.32 KB
/
cd_knn_gl_MAB_test.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
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import networkx as nx
import os
os.chdir('C:/Kaige_Research/Graph Learning/graph_learning_code/')
from collections import Counter
import datetime
from synthetic_data import *
from utils import *
from gl_MAB import GL_MAB
from knn_MAB import KNN_MAB
from cd_MAB import CD_MAB
from linucb_MAB import LINUCB_MAB
from sklearn.metrics.pairwise import rbf_kernel
from pygsp import graphs, plotting, filters
import pyunlocbox
import seaborn as sns
sns.set_style('white')
path='C:/Kaige_Research/Graph Learning/graph_learning_code/results/'
timeRun = datetime.datetime.now().strftime('_%m_%d_%H_%M_%S')
user_num=50
item_num=1000
dimension=10
item_pool_size=20
cluster_num=5
cluster_std=0.25
noise_scale=0.25
K=10
gl_alpha=1
gl_beta=0.2
gl_theta=0.01
gl_step_size=0.5
jump_step=50
alpha=0.05
iteration=5000
newpath=path+'user_num_%s_noise_scale_%s_cluster_std_%s'%(user_num, noise_scale, cluster_std)+str(timeRun)+'/'
if not os.path.exists(newpath):
os.makedirs(newpath)
noisy_signal, true_signal, item_features, true_user_features, true_label=blob_data(user_num, item_num, dimension, cluster_num, cluster_std, noise_scale)
plt.plot(noisy_signal[0], label='noisy')
plt.plot(true_signal[0], label='true')
plt.legend(loc=2)
plt.ylabel('signal', fontsize=12)
plt.xlabel('node index', fontsize=12)
plt.savefig(newpath+'true_vs_noisy_signal_%s'%(noise_scale)+'.png', dpi=100)
plt.clf()
true_adj=rbf_kernel(true_user_features)
np.fill_diagonal(true_adj, 0)
true_adj=norm_W(true_adj, user_num)
user_pool=generate_all_random_users(iteration, user_num)
item_pools=generate_all_article_pool(iteration, item_pool_size, item_num)
#########################
linucb_mab=LINUCB_MAB(user_num, item_num, dimension, item_pool_size, alpha, true_user_features=true_user_features, true_graph=true_adj)
linucb_cum_regret, linucb_user_f, linucb_error=linucb_mab.run(user_pool, item_pools, item_features, noisy_signal, true_signal, iteration)
cd_mab=CD_MAB(user_num, item_num, dimension, item_pool_size, alpha, K=5, jump_step=10, true_user_features=true_user_features, true_graph=true_adj)
cd_cum_regret, cd_adj, cd_user_f, cd_cluster_f, cd_error, cd_graph_error, cd_cluster, cd_cluster_score=cd_mab.run(user_pool, item_pools, item_features, noisy_signal,true_signal, iteration, true_label)
knn_mab=KNN_MAB(user_num, item_num, dimension, item_pool_size,alpha, K=5, jump_step=10, mode=1, true_user_features=true_user_features, true_graph=true_adj)
knn_cum_regret, knn_adj,knn_user_f, knn_error, knn_graph_error, knn_denoised_signal=knn_mab.run(user_pool, item_pools, item_features, noisy_signal, true_signal, iteration)
gl_mab=GL_MAB(user_num, item_num, dimension, item_pool_size, alpha, gl_alpha, gl_beta, gl_theta, gl_step_size, jump_step=jump_step, mode=2, true_user_features=true_user_features, true_graph=true_adj)
gl_cum_regret, gl_adj, gl_user_f, gl_error, gl_graph_error, gl_denoised_signal=gl_mab.run(user_pool, item_pools, item_features, noisy_signal,true_signal, iteration)
plt.plot(gl_cum_regret, label='GL')
plt.plot(knn_cum_regret, label='KNN')
plt.plot(cd_cum_regret, label='CD')
plt.plot(linucb_cum_regret, label='LINUCB')
plt.ylabel('Cumulative Regret', fontsize=12)
plt.xlabel('Time', fontsize=12)
plt.legend(loc=2)
plt.savefig(newpath+'cum_regret_%s_%s_%s'%(user_num, int(100*noise_scale), int(100*cluster_std))+'.png', dpi=100)
plt.show()
plt.plot(gl_error, label='GL')
plt.plot(knn_error, label='KNN')
plt.plot(cd_error, label='CD')
plt.plot(linucb_error, label='LINUCB')
plt.ylabel('Learning Error', fontsize=12)
plt.xlabel('Time', fontsize=12)
plt.legend(loc=1)
plt.savefig(newpath+'learning_error_%s_%s_%s'%(user_num,int(100*noise_scale), int(100*cluster_std))+'.png', dpi=100)
plt.show()
plt.plot(gl_graph_error, label='GL')
plt.plot(knn_graph_error, label='KNN')
plt.plot(cd_graph_error, label='CD')
plt.ylabel('Graph Error', fontsize=12)
plt.xlabel('Time', fontsize=12)
plt.legend(loc=1)
plt.savefig(newpath+'graph_error_%s_%s_%s'%(user_num,int(100*noise_scale), int(100*cluster_std))+'.png', dpi=100)
plt.show()
plt.plot(cd_cluster_score, label='CD')
plt.title('Cluster Score', fontsize=12)
plt.show()
pos=true_user_features
test_item=np.random.normal(size=dimension)
true_payoff=np.dot(true_user_features, test_item)
noise=np.random.normal(size=user_num)
noisy_payoff=true_payoff+noise
linucb_payoff=np.dot(linucb_user_f, test_item)
cd_payoff=np.dot(cd_cluster_f, test_item)
knn_payoff=np.dot(knn_user_f, test_item)
gl_payoff=np.dot(gl_user_f, test_item)
fig, axes=plt.subplots(3,2)
axes[0,0].scatter(pos[:,0], pos[:,1], c=true_payoff, cmap=plt.cm.jet)
axes[0,1].scatter(pos[:,0], pos[:,1], c=noisy_payoff, cmap=plt.cm.jet)
axes[1,0].scatter(pos[:,0], pos[:,1], c=linucb_payoff, cmap=plt.cm.jet)
axes[1,1].scatter(pos[:,0], pos[:,1], c=cd_payoff, cmap=plt.cm.jet)
axes[2,0].scatter(pos[:,0], pos[:,1], c=knn_payoff, cmap=plt.cm.jet)
axes[2,1].scatter(pos[:,0], pos[:,1], c=gl_payoff, cmap=plt.cm.jet)
axes[0,0].set_title('True signal')
axes[0,1].set_title('Noisy Signal')
axes[1,0].set_title('LINUCB Signal')
axes[1,1].set_title('CD Signal')
axes[2,0].set_title('KNN Signal')
axes[2,1].set_title('GL Signal')
plt.tight_layout()
plt.savefig(newpath+'signal_%s_%s_%s'%(user_num, int(100*noise_scale), int(100*cluster_std))+'.png', dpi=100)
plt.show()
gl_payoff2=total_variation_signal_learning(gl_adj, noisy_payoff)
fig, (ax1, ax2)=plt.subplots(2)
ax1.scatter(pos[:,0], pos[:,1], c=gl_payoff, cmap=plt.cm.jet)
ax1.set_title('GL Tikhonov Reg')
ax2.scatter(pos[:,0], pos[:,1], c=gl_payoff2, cmap=plt.cm.jet)
ax2.set_title('GL TV Reg')
plt.tight_layout()
plt.show()
# pos=true_user_features
# graph=create_networkx_graph(user_num, true_adj)
# edge_color=true_adj[np.triu_indices(user_num,1)]
# plt.figure(figsize=(5,5))
# nodes=nx.draw_networkx_nodes(graph, pos, node_color=noisy_signal[0], node_size=100, cmap=plt.cm.jet)
# edges=nx.draw_networkx_edges(graph, pos, width=1.0, alpha=0.1, edge_color='grey')
# plt.axis('off')
# plt.title('True Graph', fontsize=12)
# plt.show()
# pos=true_user_features
# graph=create_networkx_graph(user_num, gl_adj)
# edge_color=gl_adj[np.triu_indices(user_num,1)]
# plt.figure(figsize=(5,5))
# nodes=nx.draw_networkx_nodes(graph, pos, node_color=gl_denoised_signal[0], node_size=100, cmap=plt.cm.jet)
# edges=nx.draw_networkx_edges(graph, pos, width=1.0, alpha=0.1, edge_color='grey')
# plt.axis('off')
# plt.title('GL Graph', fontsize=12)
# plt.show()
# pos=true_user_features
# graph=create_networkx_graph(user_num, knn_adj)
# edge_color=knn_adj[np.triu_indices(user_num,1)]
# plt.figure(figsize=(5,5))
# nodes=nx.draw_networkx_nodes(graph, pos, node_color=knn_denoised_signal[0], node_size=100, cmap=plt.cm.jet)
# edges=nx.draw_networkx_edges(graph, pos, width=1.0, alpha=0.1, edge_color='grey')
# plt.axis('off')
# plt.title('KNN Graph', fontsize=12)
# plt.show()
# pos=true_user_features
# graph=create_networkx_graph(user_num,cd_adj)
# edge_color=cd_adj[np.triu_indices(user_num,1)]
# plt.figure(figsize=(5,5))
# nodes=nx.draw_networkx_nodes(graph, pos, node_color=cd_cluster, node_size=100, cmap=plt.cm.jet)
# edges=nx.draw_networkx_edges(graph, pos, width=1.0, alpha=0.1, edge_color='grey')
# plt.axis('off')
# plt.title('CD Graph', fontsize=12)
# plt.show()