-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmain_ssa_stats.py
147 lines (135 loc) · 4.97 KB
/
main_ssa_stats.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
import numpy as np
import matplotlib.pylab as plt
from itertools import combinations
import argparse
import random
import pickle
def subset_fixed_size(target, numbers, eps, subsize, errBest):
n = len(numbers)
cand = 0
indBest = np.array([np.NAN])
for ind in combinations(range(n),subsize):
inda = np.array(ind,dtype="int")
napprox = np.sum(numbers[inda])
diff = np.abs(target-napprox)
if diff < errBest:
errBest = diff
cand = napprox
indBest = inda
if diff <= eps:
break
return cand, indBest, errBest
def exhaustive(target, numbers, eps, nmax):
n = len(numbers)
err = np.abs(target)
errBest = err
cand = 0
indBest = np.array([-1])
nmax = min(nmax, n)
for k in range(nmax):
cank, indk, errk = subset_fixed_size(target, numbers, eps, k, errBest)
if errk < errBest:
errBest = errk
cand = cank
indBest = indk
if errBest <= eps:
break
return cand, indBest
def subset_fixed_size_best(target, numbers, subsize, errBest):
n = len(numbers)
cand = 0
indBest = np.array([-1])
for ind in combinations(range(n),subsize):
inda = np.array(ind,dtype="int")
napprox = np.sum(numbers[inda])
diff = np.abs(target-napprox)
if diff < errBest:
errBest = diff
cand = napprox
indBest = inda
return cand, indBest, errBest
def exhaustiveBest(target, numbers, nmax):
n = len(numbers)
err = np.abs(target)
errBest = err
cand = 0
indBest = np.array([-1])
nmax = min(nmax, n)
for k in range(nmax):
cank, indk, errk = subset_fixed_size_best(target, numbers, k, errBest)
if errk < errBest:
errBest = errk
cand = cank
indBest = indk
return cand, indBest
def expC(n, eps, samples, subset):
rep = samples
err = np.zeros(rep)
ii = np.zeros(rep)
for i in range(rep):
target = np.random.uniform(-1,1,1)
cand, ind = exhaustive(target, np.random.uniform(-1,1,n), eps, subset)
err[i] = np.abs(cand-target)
ii[i] = len(ind)
delta = np.sum(err > eps)/rep
if delta > 0:
C = n/(-np.log(min(eps, delta)))
else:
C = 1
return C, delta, err, ii
def main():
global args
parser = argparse.ArgumentParser(description='Constructing convolutional lottery tickets (LTs) from target models.')
parser.add_argument('--error', type=float, default=0.01, metavar='eps', help='Allowed approximation error for each target parameter (default=0.01).')
parser.add_argument('--rep', type=int, default=100000, metavar='nbrRep',
help='Number of independent repetitions of LT construction for a given target (default: 5).')
parser.add_argument('--ssa_size', type=int, default=15, metavar='rho',
help='Size of base set for subset sum approximation (and thus multiplicity of neuron construction in LT).')
parser.add_argument('--construct', type=str, default="L+1",
help='Construction method: L+1 or 2L.')
parser.add_argument('--sub', type=int, default=3,
help='Maximum considered subset size.')
parser.add_argument('--seed', type=int, default=1, help='Random seed (default=1).')
args = parser.parse_args()
random.seed(args.seed)
rep = args.rep
err = np.zeros(rep)
ii = np.zeros(rep)
n = args.ssa_size
eps = args.error
if args.construct == "L+1":
for i in range(rep):
target = np.random.uniform(-1,1,1)
cand, ind = exhaustive(target, np.random.uniform(-1,1,n), eps, args.sub) #exhaustiveBest(target, np.random.uniform(-1,1,n), 15)
err[i] = np.abs(cand-target)
ii[i] = len(ind)
if ind[0] == (-1):
ii[i] = 0
if (i%1000 == 0) and (i>1):
print(i)
print(np.mean(err[(i-1000):i]))
print("Mean error:")
print(np.mean(err))
print("Max error:")
print(np.max(err))
with open('./Subset_sum_stats/subset_stats_n_'+str(n)+'_sub_'+str(args.sub)+'_eps_'+str(eps), 'wb') as f:
pickle.dump([err,ii], f)
else:
for i in range(rep):
target = np.random.uniform(-1,1,1)
cand, ind = exhaustive(target, np.random.uniform(-1,1,n)*np.random.uniform(-1,1,n), eps, args.sub)
err[i] = np.abs(cand-target)
ii[i] = len(ind)
if ind[0] == (-1):
ii[i] = 0
if (i%1000 == 0) and (i>1):
print(i)
print(np.mean(err[(i-1000):i]))
print("Mean error:")
print(np.mean(err))
print("Max error:")
print(np.max(err))
with open('./Subset_sum_stats/subset_2l_stats_n_'+str(n)+'_sub_'+str(args.sub)+'_eps_'+str(eps), 'wb') as f:
pickle.dump([err,ii], f)
if __name__ == '__main__':
main()