/
main_onlocal_pll.py
118 lines (97 loc) · 3.3 KB
/
main_onlocal_pll.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
# FDR framework
import numpy as np
np.set_printoptions(precision = 4)
import os
import logging, argparse
import time
import itertools
from datetime import datetime
import StringIO
# import
import exp_new
import exp_new_punif
from importme import *
from ipyparallel import Client
def main():
# Get arguments
ny_data = args.ny_data
mu_style = args.mu_style
dist_style = args.dist_style
numrunrange = [25]
#### Set experimental settings in paper
pi1range = [0.4]
hyp_style = 1
top_arms = 10
alpha0 = 0.1 # generally 0.1
FDRrange = [0]
algrange = [0,1] # both 0: best-arm MAB and 1: Uniform sampling
# Power or FDR plots
if args.power_plot == 1:
punif = 0
else:
punif = 1
if (dist_style == 1):
mu_gap = 3
mu_best = 8
num_hyp = 500
# armrange = np.arange(10, 121, 10)
# truncrange = np.arange(100, 801, 100)
armrange = np.arange(10, 41, 10)
truncrange = np.arange(100, 301, 100)
if punif == 1:
armrange = [30]
algrange = [0]
truncrange = [200]
pi1range = np.arange(0.1,1,0.1)
FDRrange = [0, 3, 5]
num_hyp = 1000
elif (dist_style == 0):
mu_gap = 0.3
mu_best = 0.7
num_hyp = 50
# armrange = np.arange(5, 36, 5)
# truncrange = np.arange(5000, 25001, 1000)
truncrange = np.arange(5000, 10001, 5000)
armrange = np.arange(5, 11, 5)
# Overwrite all settings if New Yorker data
if (ny_data == 1):
dist_style = 0
mu_gap = 0
mu_style = 0
mu_best = 0
num_hyp = 30
top_arms = 5
# armrange = np.arange(10, 81, 10)
truncrange = np.arange(13000, 13001, 1000)
armrange = np.arange(10, 21, 10)
##### Run for loop
forrange = list(itertools.product(armrange, algrange, truncrange,numrunrange, pi1range, FDRrange))
def wrapper(i):
(no_arms, algnum, trunctime, numrun, pi1, FDR) = forrange[i]
if punif == 0:
exp_new.run_single(dist_style, mu_gap, mu_style, hyp_style, pi1, no_arms, num_hyp, 0,0, top_arms, alpha0,[trunctime], FDR, numrun, mu_best, algnum, verbose = 0)
else:
exp_new_punif.run_single(dist_style, mu_gap, mu_style, hyp_style, pi1, no_arms, num_hyp, 0,0, top_arms, alpha0,[trunctime], FDR, numrun, mu_best, algnum, 1, verbose = 0)
print("time for %d, %d, %d: %d" % (no_arms, algnum, trunctime, time.time() - time1))
return True
# Parallelization
c = Client()
c.ids
dview = c[:]
dview.block = True
lview = c.load_balanced_view()
lview.block = True
dview.execute('import numpy as np')
dview.execute('import exp_new')
mydict = dict(forrange = forrange, hyp_style = hyp_style, num_hyp = num_hyp, alpha0 = alpha0, dist_style = dist_style)
dview.push(mydict)
lview.map(wrapper, range(len(forrange)))
if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument('--mu-style',type=int, default = 2)
parser.add_argument('--dist-style', type=int, default = 1)
parser.add_argument('--ny-data', type=int, default = 0)
parser.add_argument('--power-plot', type=int, default = 1)
args = parser.parse_args()
logging.info(args)
main()