/
streakRegressionExample.py
215 lines (190 loc) · 8.96 KB
/
streakRegressionExample.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
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
import numpy as np
import pandas as pd
import sys
import csv
from scipy import io as sio
from scipy import stats
from sklearn import svm, preprocessing, base, datasets, feature_selection, linear_model, decomposition
from os import listdir
from os.path import isfile, join, exists
import argparse
from itertools import combinations,product,chain,combinations_with_replacement
from functools import reduce
import random
from streakFunctions import *
from joblib import Parallel, delayed
from joblib import load, dump
from time import time
import load_data
import streakFunctions as streak
def runStreamingLogisticMain(numTrain=800,K=80,numIt=10,regParam=0,streamTolerance=0.5,logFlag=1):
if logFlag:
resultsFilename = 'resultsReg.txt'
if not exists(resultsFilename):
#initialize results file with table
fptr = open(resultsFilename,'w')
fptr.write('data' + '\t' + 'Ntr' + '\t' + 'p' + '\t' + 'K' '\t' + 'alg' + '\t' + 'tol' '\t' +
'llhd' + '\t' + 'genScore' + '\t' + 'time' + '\t' + 'fevals' + '\n')
fptr.close()
# #select a dataset
# saveString = 'phis'
# dataString = 'phishL'
# p = 68
# pdim = p
# # m = regParam
# runVerbose = True
# Ntotal = 2*numTrain
# saveString = saveString + "_Ntr" + str(numTrain) + "_tol" + str(streamTolerance) + "_K" + str(K) + "_numIt" + str(numIt)
saveString = 'phispair'
dataString = 'phishP'
p = 68
pdim = (p**2 + 3*p)/2
# m = regParam
runVerbose = True
Ntotal = 2*numTrain
saveString = saveString + "_Ntr" + str(numTrain) + "_tol" + str(streamTolerance) + "_K" + str(K) + "_numIt" + str(numIt)
if saveString[0:4] in ['rcv1','leuk','phis']:
if runVerbose:
llStrNorms = np.zeros((pdim+1,numIt))
else:
llStrNorms = np.zeros((K,numIt))
scoreStrs = np.zeros_like(llStrNorms)
timeStrs = np.zeros(numIt)
evalStrs = np.zeros_like(timeStrs)
#initialize arrays, print or store?
if runVerbose:
llRandNorms = np.zeros((pdim+1,numIt))
else:
llRandNorms = np.zeros((K,numIt))
llLocalNorms = np.zeros_like(llRandNorms)
scoreRands = np.zeros_like(llRandNorms)
scoreLocals = np.zeros_like(llRandNorms)
timeLocals = np.zeros(numIt)
timeRands = np.zeros_like(timeLocals)
evalLocals = np.zeros_like(timeLocals)
evalRands = np.zeros_like(timeLocals)
for it in np.arange(numIt):
print "ITERATION: %d, N_training=%d, dim=%d, K=%d, tolerance=%f, regularizer=%f" % (it,numTrain,p,K,streamTolerance,regParam)
#read dataset (downloaded from https://www.csie.ntu.edu.tw/~cjlin/libsvmtools/datasets/binary.html)
if saveString[0:4] == 'phis':
X_tot,y_tot = load_data.loadFromLibsvm('phishing.txt',Ntotal,p)
# random permutation of data samples
nperm = np.random.permutation(Ntotal)
X_train = X_tot[nperm[:numTrain]]
y_train = y_tot[nperm[:numTrain]]
X_test = X_tot[nperm[numTrain:]]
y_test = y_tot[nperm[numTrain:]]
print "loaded train and test sets (each size %d) from phishing dataset" % numTrain
# X_train = X_train - np.mean(X_train,0)
# X_test = X_test - np.mean(X_test,0)
# y_train = 0.5*(y_train + 1)
# y_test = 0.5*(y_test + 1)
elif saveString[0:4] == 'leuk':
X_train,y_train = load_data.loadFromLibsvm('leu',numTrain,p)
X_test,y_test = load_data.loadFromLibsvm('leu.t',numTrain,p)
print "loaded train and test sets (each size %d) from leukemia dataset" % numTrain
X_train = X_train - np.mean(X_train,0)
X_test = X_test - np.mean(X_test,0)
y_train = 0.5*(y_train + 1)
y_test = 0.5*(y_test + 1)
elif saveString[0:4] == 'rcv1':
X_train,y_train = load_data.loadFromLibsvm('rcv1_train.binary',numTrain,p)
X_test,y_test = load_data.loadFromLibsvm('rcv1_test.binary',numTrain,p)
print "loaded train and test sets (each size %d) from rcv1 binary dataset" % numTrain
X_train = X_train - np.mean(X_train,0)
X_test = X_test - np.mean(X_test,0)
#map the y's that have from -1 to 0
#map from 1 -1 to 1 0 1/2 (x+1)
#random permutation every time? no, this is taken elsewhere
# pperm = np.random.permutation(p)
# X_train = X_train[:,pperm]
# X_test = X_test[:,pperm]
y_train = 0.5*(y_train + 1)
y_test = 0.5*(y_test + 1)
llNull = streak.getLogLikelihood(np.zeros(p),np.zeros(p),y_train,regParam)
if regParam == 0:
regVal = 1.0/Cinf
else:
regVal = regParam
if saveString[0:4] in ['rcv1','leuk','gise','sona','phis']:
if saveString[0:7] in ['rcv1pai','leukpai','gisepai','sonapai','phispai']:
print "Streaming Pairs"
streamIter = chain(range(p),combinations_with_replacement(range(p),2))
randomStreamIter = streak.random_permutation(streamIter)
else:
print "Streaming"
randomStreamIter = streak.random_permutation(range(p))
t1 = time()
bsupS,llStr,scoreStrs[:,it],evalStrs[it] = streak.solveStreamingRegression(K,streamTolerance,
X_train,y_train,X_test,y_test,llNull,regVal,randomStreamIter,'logistic')
t2 = time()
timeStrs[it] = t2-t1
print "streaming finished in %f" % (t2-t1)
if logFlag:
with open(resultsFilename,'a') as lptr:
lptr.write('\t'.join([dataString,str(numTrain),str(p),str(K),'stream',str(streamTolerance),
str(llStr[-1]),str(scoreStrs[-1,it]),str(timeStrs[it]),str(int(evalStrs[it]))]) + '\n')
print "Random"
t1 = time()
if runVerbose:
bsupR,llRand,scoreRands[1:(K+1),it],evalRands[it] = streak.solveRandomLogistic(K,X_train,y_train,X_test,y_test,
llNull,regVal,randomStreamIter)
#TODO: pad with zeros?
llRand = np.concatenate(( np.array([0.]),llRand,llRand[K-1]*np.ones((pdim-K)) ))
scoreRands[(K+1):,it] = scoreRands[K,it]
else:
bsupR,llRand,scoreRands[:,it],evalRands[it] = streak.solveRandomLogistic(K,X_train,y_train,X_test,y_test,
llNull,regVal,randomStreamIter)
t2 = time()
timeRands[it] = t2-t1
# print llRand,scoreRands
print "random finished in %f" % (t2-t1)
if logFlag:
with open(resultsFilename,'a') as lptr:
#give random subset a tolerance of '2' so we can sort by this field later
lptr.write('\t'.join([dataString,str(numTrain),str(p),str(K),'random','2',str(llRand[-1]),
str(scoreRands[-1,it]),str(timeRands[it]),str(int(evalRands[it]))]) + '\n')
if saveString[0:7] in ['rcv1pai','leukpai','gisepai','sonapai','phispai']:
print "Local Search (and generating all pairwise features)"
else:
print "Local Search"
t1 = time()
bsupL,llLocal,scoreLocals[:,it],evalLocals[it] = streak.solveLocalSearchLogistic(K,X_train,y_train,X_test,y_test,
llNull,regVal,randomStreamIter).getStats()
t2 = time()
timeLocals[it] = t2-t1
if logFlag:
with open(resultsFilename,'a') as lptr:
#give local search a tolerance of '-1' so we can sort by this field later
lptr.write('\t'.join([dataString,str(numTrain),str(p),str(K),'localsearch','-1',str(llLocal[-1]),
str(scoreLocals[-1,it]),str(timeLocals[it]),str(int(evalLocals[it]))]) + '\n')
print "localsearch finished in %f" % (t2-t1)
else:
raise ValueError("unsupported dataset")
llStrNorms[:,it] = llStr
llLocalNorms[:,it] = llLocal
llRandNorms[:,it] = llRand
if __name__ == '__main__':
argsList = []
loggingFlag = 1
# nVec = [50]
nVec = [2000]
tolVec = [0.75,0.1]
kVec = [20]
# kVec = [20,40,80]
regVec = [0.0]
for n in nVec:
for reg in regVec:
for k in kVec:
for tol in tolVec:
argsList.append({
'numTrain': n,
'K': k,
'regParam': reg,
'streamTolerance': tol,
'logFlag': loggingFlag,
'numIt': 2})
#run in parallel
njobs = 1 #only 1 core
# njobs = -3 #use all but 2 cores
r = Parallel(n_jobs=njobs)(delayed(runStreamingLogisticMain)(**args) for args in argsList)