/
Rgrad_old_step.py
executable file
·191 lines (163 loc) · 5.42 KB
/
Rgrad_old_step.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
#-*- coding: utf-8 -*-
"""
Created on Mon Mar 11 18:49:56 2013
@author: rami999999999
"""
import pdb
import gc
import numpy as np
import scipy.sparse as sparse
import iteracoes as it
import cria_dados
import Rfechado as RR
def grad(X,Y,W0,Xteste,Yteste,Xdev,Ydev,arg_lamb, max_iter,fixo): #funcao que implementa o algoritmo gradient descent, Y nao normalizado
from copy import copy, deepcopy
w_old=W0
import numpy as np
import scipy.sparse as sparse
import iteracoes as it
import cria_dados
import Rfechado as RR
if arg_lamb == False:
#lambs=[10**x for x in xrange(-2,10)]
lambs=[10]
xg=[1]
else:
lambs=[arg_lamb]
xg=[np.log10(arg_lamb)]
#max_iter =30000000
gc.collect()
#############SIGMA######################
# #
#********Para lambda>1 funciona********#
sigma=1 #
sigma2=0.0000001 #
# #
#********Para lambda <1 funciona*******#
#sigma=0.0001 #
#sigma2=0.000001 #
# #
########################################
errorg=[]
errortg=[]
final=[-1,0,0]
for lamb in lambs:
print "A iniciar iteracoes para lambda=", lamb
xx=[]
y_old=0
yy=[]
w_old=W0
i=0
for i in xrange(max_iter):
if fixo==True:
step_size=1e-10
else:
step=1e-9
step_size=step/np.sqrt(i+1)
error=(X*w_old)-Y
grad1=it.get_gradient(error,X,w_old,lamb)
w_new = w_old - (step_size * grad1)
w_new=w_new.tocsr()
error=(X*w_new)-Y
#dif=w_new-w_old
#dif=dif.transpose()*dif
y_new=it.get_func(error,w_new,lamb) #funcao de erro
'''
count=0
if i!=0:
while y_new>=y_old-sigma2*alpha*dif[0,0]:
#print "A diminuir step:",i
step_size=step_size/2
w_new = w_old - (step_size * grad1)
w_new=w_new.tocsr()
error=(X*w_new)-Y
dif=w_new-w_old
dif=dif.transpose()*dif
y_new=it.get_func(error,w_new,lamb) #funcao de erro
count=count+1
#y_new=y_new[0,0]
if count==1000:
break
if count ==1000:
#print "****A SAIR****\nO sparsa encontrou o minimo"
break
'''
y_old=y_new
#print y_new
w_old = w_new
#print "y_new:",y_new
yy.append(y_new)
xx.append(i)
#yyy.append(((grad1.transpose()*grad1).data)[0])
#i=i+1
#pdb.set_trace()
error=RR.erro(Xdev,Ydev,w_new)
errorg.append(error)
errortg.append(RR.erro(Xteste,Yteste,w_new))
if final[0]>error or final[0]==-1:
final[0]=error
final[1]=lamb
final[2]=w_new
yyg=deepcopy(yy)
xxg=deepcopy(xx)
print final[1]
'''
import os
import pylab
os.environ['DISPLAY']
import matplotlib.pyplot as plt
plt.figure(1)
plt.title("n = 1e-8")
plt.plot(xxg,yyg,"b")
plt.plot(xxg,yyg,"b",xxg,[custo for cenas in xrange(len(xxg))],"r--")
pylab.ylim([custo-1e90,custo+8e90])
'''
'''
plt.subplot(311)
plt.title("Funcao de custo ao longo das iteracoes")
plt.plot(xxg,yyg,"b")
#plt.plot(xxg,yyg,"b",xxg,[9.42247e+15 for cenas in xrange(len(xxg))],"r--")
#pylab.ylim([4.082103e+15,15.082103e+15])
plt.subplot(312)
#plt.title("Modulo do gradiente")
#plt.plot(xx,yyy,"b")
plt.title("Erro Test para os varios lambdas")
plt.plot(xg,errortg,"b",xg,errortg,"ro")
plt.subplot(313)
plt.title("Erro Dev para os varios lambdas")
plt.plot(xg,errorg,"b",xg,errorg,"ro")
'''
print "FUNCAO DE CUSTO"
#print yyg[len(yyg)-1]
#plt.show()
return final[2],yyg,xxg
def main():
media=0
maximo=1
f="../le_ficheiro/someta"
dictionary,total,y=cria_dados.read_output(f+"train.txt")
X,Y=cria_dados.criaXY(dictionary,total,y,False)
#X,total=cria_dados.delcomun(X,total)
dictionary,temp,y=cria_dados.read_output(f+"test.txt")
Xteste,Yteste=cria_dados.criaXY(dictionary,total,y,False)
dictionary,temp,y=cria_dados.read_output(f+"dev.txt")
Xdev,Ydev=cria_dados.criaXY(dictionary,total,y,False)
dictionaty=[]
y=[]
gc.collect()
vec=sparse.csr_matrix([0 for i in xrange(X.shape[1])])
vec=vec.transpose()
W=grad(X,Y,vec,Xteste,Yteste,Xdev,Ydev)
print "W:"
print W.todense()
print "------------------------------"
print "MSE(teste):\n",RR.erro(f+"test.txt",W,total,media,maximo)
print "------------------------------"
print "------------------------------"
#print "MSE(train):\n",erro(f+"train.txt",W,total)
#print "------------------------------"
print "MSE(dev):\n",RR.erro(f+"dev.txt",W,total,media,maximo)
#error=(X*W)-Y
#print "Funcao de custo:",get_func(error,W,10)
if __name__=="__main__":
main()