/
test_approx_heuristic.py
175 lines (153 loc) · 9.12 KB
/
test_approx_heuristic.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
from ic_bfs_eval import *
from generate_seeds import LoadNodesFromFile, generateSeedFiles
import subprocess
from link_server import LinkServerCP
from ic_bfs_eval import EstimateInfluence
from bfs_seq import sequential_estimation
from time import time
import argparse
import numpy as np
from math import log, ceil
import gc
from math import log
from common_tools import *
from random import sample,randint, seed
parser = argparse.ArgumentParser()
parser.add_argument('-csv', type = str, #required,
default = 'input/datasets/Wiki-Vote_stripped.txt', help="Name of source csv file (if we generate a new Link Server" )
parser.add_argument('-delim', type = int, default = 0,
help = '-delim : which delimiter (for edge-tuples) 0-tab, 1-space,2-comma. Default: 0')
parser.add_argument('-samples', type = int, default = 5,
help = '-samples : number of samples of given seed set size. Default: 10')
parser.add_argument('-cores', type = int, default = 50,
help = '-cores : number of cores of to use. Default: 50')
parser.add_argument('-undirected', type = int, default = 0,
help = '-undirected : is the graph undirected')
parser.add_argument('-output', type = str, default = 'out')
parser.add_argument('-dataset', type = str, default = 'input/datasets/wiki')
parser.add_argument('-init_samples', type = int, default = 5,
help = '-init_samples : number of samples to take for initial tau value')
parser.add_argument('-iter_samples', type = int, default = 5,
help = '-guess_samples : number of samples to take for initial tau value')
parser.add_argument('-max_k', type = float, default = .01,
help = '-max_k : maximum value for k/n. Default: .5')
parser.add_argument('-min_k', type = float, default = .001,
help = '-min_k : minimum value for k/n. Default: .05')
parser.add_argument('-k_mode', type = int, default = 0,
help = '-k_mode : 0 -- take fractional values, 1 -- min_k, max_k, k_steps are integral')
parser.add_argument('-k_step', type = float, default = .001,
help = '-k_step : .05')
parser.add_argument('-prob_method', type = int, default = 3)
parser.add_argument('-prob', type= str, default = '')
parser.add_argument('-tau_scale', type = float, default = 0.1,
help = '-tau_scale : scaling factor for tau. Default: 0.1')
def runApproxHeuristic(dataset, seeds_fname, tau_scale, cores, init_samples, iter_samples, nCycles_link_server):
res_fname = "res" + str(randint(1, 1000))
perf_fname = "perf" + str(randint(1, 1000))
subprocess.Popen("perf stat -x, -o %s python seq_estimation.py -dataset %s -seeds_file %s -results_file %s -output_mode 1 -nSamples %d -cores %d"%\
(perf_fname, dataset, seeds_fname, res_fname, init_samples, parameters.cores),\
shell=True,stdout=subprocess.PIPE).stdout.read()
init_tau = float(open(res_fname, 'r').readline().strip()) / (1 - tau_scale)
print "Initial tau value set to: ", init_tau
num_cycles_approx = getNumCycles(perf_fname) - nCycles_link_server
removeFile(perf_fname)
removeFile(res_fname)
subprocess.Popen("perf stat -x, -o %s python ic_bfs_eval.py -tau_scale %.3f -dataset %s -res_fname %s -seeds %s -output_mode 2 -cores %d -init_tau %.3f -iter_samples %d"%\
(perf_fname, tau_scale, dataset, res_fname, seeds_fname, cores, init_tau, iter_samples), \
shell = True, stdout = subprocess.PIPE).stdout.read()
estimate_approx = float(open(res_fname, 'r').readline().strip())
print "Estimated value: ", estimate_approx
num_cycles_approx += getNumCycles(perf_fname) - nCycles_link_server
removeFile(res_fname)
removeFile(perf_fname)
return (estimate_approx, num_cycles_approx)
def runVanilla(dataset, seeds_fname, nSamples, cores):
perf_fname = "perf" + str(randint(1,1000))
res_fname = "res" + str(randint(1,1000))
subprocess.Popen("perf stat -x, -o %s python seq_estimation.py -dataset %s -seeds_file %s -results_file %s -output_mode 1 -nSamples %d -cores %d"%\
(perf_fname, dataset, seeds_fname, res_fname, nSamples, cores),\
shell=True,stdout=subprocess.PIPE).stdout.read()
num_cycles_seq = getNumCycles(perf_fname) - nCycles_link_server
print "dataset: ", dataset
print "res_fname: ", res_fname
print "dataset: ", dataset
print "seeds file: ", seeds_fname
estimate_seq_capped = float(open(res_fname, 'r').readline().strip())
removeFile(perf_fname)
removeFile(res_fname)
return estimate_seq_capped, num_cycles_seq
if __name__ == "__main__":
seed()
gc.enable()
parameters = parser.parse_args()
min_k, max_k, k_step, nSamples, edges_csv, delim_option, output, dataset, tau_scale, init_samples, iter_samples = \
parameters.min_k, parameters.max_k, parameters.k_step, parameters.samples, parameters.csv, parameters.delim, \
parameters.output, parameters.dataset, parameters.tau_scale, parameters.init_samples, parameters.iter_samples
delim_dict = {0 : '\t', 1 : ' ', 2 : ','}
delimeter = delim_dict[delim_option]
results_dir = 'experiments/results/approx-heuristic1/'
bUndirected = True if parameters.undirected == 1 else False
bfs_method = 'seq'
start_time = time()
if parameters.prob_method == 2:
if parameters.prob == '':
prob_values = [0.001, 0.0001]
else:
l = parameters.prob.split(',')
prob_values = [float(x) for x in l]
elif parameters.prob_method == 3:
if parameters.prob == '':
prob_values = [0,.01]
else:
l = parameters.prob.split(',')
assert len(l) == 2
prob_values = [float(x) for x in l]
print "Probability method: %d, probability values: %s"%(parameters.prob_method, prob_values)
L=LinkServerCP(dataset, edges_csv, create_new=True, prob_method=parameters.prob_method, prob=prob_values, delim=delimeter, undirected = bUndirected)
print "Number of seed sets: ", nSamples
n = L.getNumNodes()
if parameters.k_mode == 0:
min_k, max_k, k_step = int(min_k * n), int(max_k * n), int(k_step * n)
perf_csv_fname = "perf_out" + str(randint(1,1000))
nBFS_samples_theoretic = n * log(n,2)
nBFS_samples = 1000 #nBFS_samples_theoretic
results_file = results_dir + dataset +\
'-approx-heuristic-prob_method-%d-k_min-%.4f-k_max-%.4f-tau_scale-%.3f-samples-%d-bfs_samples-%d-init_samples-%d-iter_samples-%d'%\
(parameters.prob_method, min_k, max_k, tau_scale, nSamples, nBFS_samples, init_samples, iter_samples)
removeFile(results_file)
# Record time to load link-server file
subprocess.Popen("perf stat -x, -o %s python load_link_server.py -cp %s"%\
(perf_csv_fname, dataset), shell = True, stdout = subprocess.PIPE).stdout.read()
nCycles_link_server = getNumCycles(perf_csv_fname)
print "Number of cycles for loading link server: ", nCycles_link_server
removeFile(perf_csv_fname)
for k in xrange(int(min_k),int(max_k + 1), int(k_step)):
print "k = ", k
seeds_fname = "%s-seeds-%d.cp"%(output, k)
for i in xrange(nSamples):
print "sample #",i
generateSeedFiles(k, k+1, 1, range(n), 1, output + "-seeds-")
seeds=cp.load(open(seeds_fname,'r'))
print "Running Vanilla with %d samples"%(nBFS_samples)
true_value, num_cycles_full = runVanilla(dataset, seeds_fname, nBFS_samples, parameters.cores)
print "Done. Number of cycles: %d"%num_cycles_full
nCycles_per_bfs = 1.*(num_cycles_full - nCycles_link_server) / nBFS_samples
print "Number of cycles per sample", nCycles_per_bfs
print "Running approximation algorithm"
approx_estimate, num_cycles_approx = runApproxHeuristic(dataset, seeds_fname, tau_scale, parameters.cores,\
parameters.init_samples, parameters.iter_samples, nCycles_link_server)
print "Number of cycles without link-server loading: ", num_cycles_approx
print "Done approximating, now running naive sequential algorithm"
nVanilla_samples = int(ceil(1. * num_cycles_approx / nCycles_per_bfs))
print "Running Vanilla for %s samples"%nVanilla_samples
seq_estimate, num_cycles_seq = runVanilla(dataset, seeds_fname, nVanilla_samples, parameters.cores)
num_cycles_seq = num_cycles_seq - nCycles_link_server
print "Done running capped sequential algorithm"
print "Number of cycles: ", num_cycles_seq
print "Ratio of # of cycles_approx to # of cycles_vanilla_capped : ", 1.*num_cycles_approx / num_cycles_seq
removeFile(seeds_fname)
f = open(results_file, 'a')
f.write('%.4f\t%.6f\t%.6f\t%.6f\n'%(1.*k/n if parameters.prob_method==0 else float(k), true_value, approx_estimate, seq_estimate))
f.close()
removeFile(dataset)
print "Total runtime: ", (time() - start_time)/60.0