-
Notifications
You must be signed in to change notification settings - Fork 2
/
testSocialBandits.py
executable file
·158 lines (132 loc) · 6.75 KB
/
testSocialBandits.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
#!/usr/bin/env python
# -*- coding: utf-8 -*-
import numpy as np
import pandas as pd
from SocialBandits import *
import matplotlib
#not using the X display
matplotlib.use('Agg',force=True)
import matplotlib.pyplot as plt
def testSocialBandits(P, U0, n, d, H, M, alpha, sigma, lam, delta, scale,\
finiteSet, stochastic,infinite):
if finiteSet:
BanditStrategies = ['LinOptV1', 'RegressionLinREL1FiniteSet', \
'LinREL2FiniteSet', 'LinREL1FiniteSet',\
'RandomBanditFiniteSet']
figname = 'finite_syncmp_n%d_d%d_h%d_m%d_a%f_s%f_d%f_scale%f'\
%(n,d,H,M,alpha,sigma,delta,scale)
else:
BanditStrategies = ['LinOptV1', 'RegressionLinREL1L2Ball',\
'LinREL1L2Ball', 'LinUCB','RandomBanditL2Ball']
figname = 'l2ball_syncmp_n%d_d%d_h%d_m%d_a%f_s%f_d%f_scale%f'\
%(n,d,H,M,alpha,sigma,delta,scale)
rewards = np.zeros((H, len(BanditStrategies)))
regrets = []
#fake bandit class for generating the finite set
sb1 = SocialBandit(P,U0)
sb1.generateFiniteSet(M)
fst = sb1.getFiniteSet()
for i, strategy in enumerate(BanditStrategies):
if stochastic: strategy = 'Stochastic'+strategy
elif infinite: strategy = 'Infinite'+strategy
print '\t%s'%strategy
BanditClass = eval(strategy)
if "LinREL" in strategy:
sb = BanditClass(P, U0, alpha, sigma, lam, delta, scale)
else:
sb = BanditClass(P, U0, alpha, sigma, lam)
sb.setFiniteSet(fst, M)
rewards[:,i] = sb.run(H)
if i > 0:
regret = np.cumsum(rewards[:,0] - rewards[:,i])
regrets.append({'bandit': strategy, 'regret': regret})
plt.plot(regret, label=strategy)
plt.xlabel('Horizon (time step)')
plt.ylabel('Cumulative Regret')
plt.legend(loc='upper left')
plt.savefig(figname+'.pdf',format='pdf')
#plt.show()
df = pd.DataFrame(regrets)
df.to_csv('%s.csv'%figname) # it should be renamed
def testSocialBanditsVariants(P, U0, n, d, H, M, alpha, sigma, lam, delta,\
scale, finiteSet, stochastic,infinite):
if finiteSet:
BanditStrategies = ['LinOptV1', 'LinREL1FiniteSet',\
'InfiniteLinREL1FiniteSet',\
'StochasticLinREL1FiniteSet']
figname = 'linrel1_finite_syncmp_%d_d%d_h%d_m%d_a%f_s%f_d%f_scale%f'\
%(n,d,H,M,alpha,sigma,delta,scale)
else:
BanditStrategies = ['LinOptV1', 'LinREL1L2Ball',\
'InfiniteLinREL1L2Ball',\
'StochasticLinREL1L2Ball']
figname = 'linrel1_l2_syncmp_variants_n%d_d%d_h%d_m%d_a%f_s%f_d%f_scale%f'\
%(n,d,H,M,alpha,sigma,delta,scale)
rewards = np.zeros((H, len(BanditStrategies)))
regrets = []
#fake bandit class for generating the finite set
sb1 = SocialBandit(P,U0)
sb1.generateFiniteSet(M)
fst = sb1.getFiniteSet()
for i, strategy in enumerate(BanditStrategies):
if stochastic: strategy = 'Stochastic'+strategy
elif infinite: strategy = 'Infinite'+strategy
print '\t%s'%strategy
BanditClass = eval(strategy)
if "LinREL" in strategy:
sb = BanditClass(P, U0, alpha, sigma, lam, delta, scale)
else:
sb = BanditClass(P, U0, alpha, sigma, lam)
sb.setFiniteSet(fst, M)
rewards[:,i] = sb.run(H)
if i > 0:
regret = np.cumsum(rewards[:,0] - rewards[:,i])
regrets.append({'bandit': strategy, 'regret': regret})
plt.plot(regret, label=strategy)
plt.xlabel('Horizon (time step)')
plt.ylabel('Cumulative Regret')
plt.legend(loc='upper left')
plt.savefig(figname+'.pdf',format='pdf')
#plt.show()
df = pd.DataFrame(regrets)
df.to_csv('%s.csv'%figname) # it should be renamed
if __name__=="__main__":
parser = argparse.ArgumentParser(description = 'Social Bandit Test Runner',formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument('--n', default=100, type=int,help ="Number of users")
parser.add_argument('--t', default=500, type=int,help ="Horizon")
parser.add_argument('--d', default=10, type=int,help ="Number of dimensions")
parser.add_argument('--alpha', default=0.05, type=float, help='alpha value. beta is set to 1 - alpha')
parser.add_argument('--sigma', default=0.05, type=float, help='Standard deviation σ of noise added to responses ')
parser.add_argument('--lam', default=0.01, type=float, help='Regularization parameter λ used in ridge regression')
parser.add_argument('--delta', default=0.1, type=float, help='δ value. Used by LinREL')
parser.add_argument('--scale', default=1, type=float, help=\
'scale of the β(t) value. Used by LinREL')
parser.add_argument('--M',default=100, type=int, help='Size M of finite set. Used by all finite set strategies. ')
parser.add_argument('--maxiter',default=50, type=int, help='Maximum number of iterations')
parser.add_argument('--debug',default='INFO', help='Verbosity level',choices=['DEBUG','INFO','WARNING','ERROR','CRITICAL'])
parser.add_argument('--logfile',default='SB.log',help='Log file')
parser.set_defaults(screen_output=True)
parser.add_argument('--noscreenoutput',dest="screen_output",action='store_false',help='Suppress screen output')
parser.add_argument("--randseed",type=int,default=42,help="Random seed")
parser.add_argument('--finite_set', dest='finite_set', action='store_true', help="Finite set solution")
parser.add_argument('--stochastic', dest='stochastic', action='store_true',\
help='Solution using stochastic choice of profiles')
parser.add_argument('--infinite', dest='infinite', action='store_true',\
help='Solution using stochastic choice of profiles')
parser.add_argument('--variants', dest='variants', action='store_true',\
help='Testing all the variants')
args = parser.parse_args()
np.random.seed(args.randseed)
random.seed(args.randseed)
P = generateP(args.n)
U0 = np.random.randn(args.n,args.d)
if not args.variants:
testSocialBandits(P, U0, args.n, args.d, args.t, args.M,\
args.alpha, args.sigma, args.lam, args.delta,\
args.scale, args.finite_set, args.stochastic,\
args.infinite)
else:
testSocialBanditsVariants(P, U0, args.n, args.d, args.t, args.M,\
args.alpha, args.sigma, args.lam, args.delta,\
args.scale, args.finite_set, args.stochastic,\
args.infinite)