-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathregression.py
163 lines (119 loc) · 5.48 KB
/
regression.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
import __main__
import numpy
import pylab
import sys
from contrib.modelfit import *
import contrib.modelfit
import contrib.dd
import contrib.JanA.dataimport
from contrib.JanA.visualization import compareModelPerformanceWithRPI, showRFS
def bilinearRegression(training_inputs,training_set,sizex,sizey,num_steps,alpha,num_neurons):
laplace = laplaceBias(sizex,sizey)
(num_pres,kernel_size) = numpy.shape(training_inputs)
X = numpy.mat(training_inputs)
Y = numpy.mat(training_set)
print numpy.shape(alpha*laplace)
print numpy.shape(X.T*X)
K1 = []
K2 = []
kk1 = numpy.linalg.pinv(X.T*X + alpha*laplace) * X.T * Y
M = numpy.zeros((num_pres,kernel_size,kernel_size))
for xx in xrange(0,sizex):
for yy in xrange(0,sizey):
for x in xrange(0,sizex):
for y in xrange(0,sizey):
if( ((xx + (x-sizex/2)) < 0) or ((xx + (x-sizex/2)) >= sizex) or ((yy + (y-sizey/2)) < 0) or ((yy + (y-sizey/2)) >= sizey)):
M[:,yy*sizex+xx, y*sizex+x] = 0
else:
M[:,yy*sizex+xx, y*sizex+x] = training_inputs[:,(yy + (y-sizey/2))*sizex + (xx + (x-sizex/2))]
for n in xrange(0,num_neurons):
print 'Neuron: ', n
k1 = kk1[:,n]
for i in xrange(0,num_steps):
print 'Step: ', i
A = numpy.hstack([numpy.mat(M[:,:,i]) * k1 for i in xrange(0,sizex*sizey)])
k2 = numpy.linalg.pinv(A.T*A + alpha*laplace) * A.T * Y[:,n]
B = numpy.hstack([numpy.mat(M[:,i,:]) * k2 for i in xrange(0,sizex*sizey)])
k1 = numpy.linalg.pinv(B.T*B + alpha*laplace) * B.T * Y[:,n]
K1.append(k1)
K2.append(k2)
K1 = numpy.hstack(K1).T
K2 = numpy.hstack(K2).T
print sizex
print sizey
print shape(K1)
numpy.reshape(numpy.array(K1),(-1,sizex,sizey))
print numpy.shape(numpy.reshape(numpy.array(K1),(-1,sizex,sizey)))
showRFS(numpy.reshape(numpy.array(K1),(-1,sizex,sizey)))
release_fig('K1.png')
showRFS(numpy.reshape(numpy.array(K2),(-1,sizex,sizey)))
release_fig('K2.png')
return (K1,K2)
def bilinearModelResponse(inputs,k1,k2,sizex,sizey):
(num_pres,kernel_size) = numpy.shape(inputs)
num_neurons = numpy.shape(k1)[0]
print sizex,sizey,kernel_size
M = numpy.zeros((num_pres,kernel_size,kernel_size))
for xx in xrange(0,sizex):
for yy in xrange(0,sizey):
for x in xrange(0,sizex):
for y in xrange(0,sizey):
if( ((xx + (x-sizex/2)) < 0) or ((xx + (x-sizex/2)) >= sizex) or ((yy + (y-sizey/2)) < 0) or ((yy + (y-sizey/2)) >= sizey)):
M[:,yy*sizex+xx, y*sizex+x] = 0
else:
M[:,yy*sizex+xx, y*sizex+x] = inputs[:,(yy + (y-sizey/2))*sizex + (xx + (x-sizex/2))]
response=[]
for n in xrange(0,num_neurons):
A = numpy.hstack([numpy.mat(M[:,:,i]) * k1.T[:,n] for i in xrange(0,sizex*sizey)])
response.append(A*k2.T[:,n])
return numpy.hstack(response)
def runBilinearRegression():
d = contrib.dd.loadResults("newest_dataset.dat")
(sizex,sizey,training_inputs,training_set,validation_inputs,validation_set,ff,db_node) = sortOutLoading(d)
raw_validation_set = db_node.data["raw_validation_set"]
contrib.modelfit.save_fig_directory='/home/antolikjan/Doc/reports/Sparsness/InputContext/'
params={}
params["Bilinear"]=True
db_node = db_node.get_child(params)
params={}
params["alpha"] = __main__.__dict__.get('Alpha',0.02)
params["num_steps"] = __main__.__dict__.get('NumSteps',1)
params["num_neurons"]= __main__.__dict__.get('NumNeurons',10)
db_node = db_node.get_child(params)
kernels = bilinearRegression(training_inputs,training_set,sizex,sizey,params["num_steps"],params["alpha"],params["num_neurons"])
pred_act = bilinearModelResponse(training_inputs,kernels[0],kernels[1],sizex,sizey)
pred_val_act = bilinearModelResponse(validation_inputs,kernels[0],kernels[1],sizex,sizey)
compareModelPerformanceWithRPI(training_set[:,0:params["num_neurons"]],validation_set[:,0:params["num_neurons"]],training_inputs,validation_inputs,numpy.mat(pred_act),numpy.mat(pred_val_act),numpy.array(raw_validation_set)[:,:,0:params["num_neurons"]],'BilinearModel')
db_node.add_data("Kernels",kernels,force=True)
contrib.dd.saveResults(d,"newest_dataset.dat")
def analyseBlilinearModel():
d = contrib.dd.loadResults("newest_dataset.dat")
dataset_node = d.children[0].children[0]
training_set = dataset_node.data["training_set"]
validation_set = dataset_node.data["validation_set"]
training_inputs= dataset_node.data["training_inputs"]
validation_inputs= dataset_node.data["validation_inputs"]
raw_validation_set = dataset_node.data["raw_validation_set"]
kernels = dataset_node.children[1].children[4].data['Kernels']
K1 = kernels[0]
K2 = kernels[1]
pred_act = numpy.multiply(training_inputs*K1.T,training_inputs*K2.T)
pred_val_act = numpy.multiply(validation_inputs*K1.T,validation_inputs*K2.T)
compareModelPerformanceWithRPI(training_set,validation_set,training_inputs,validation_inputs,numpy.mat(pred_act),numpy.mat(pred_val_act),raw_validation_set,85)
def laplaceBias(sizex,sizey):
S = numpy.zeros((sizex*sizey,sizex*sizey))
for x in xrange(0,sizex):
for y in xrange(0,sizey):
norm = numpy.mat(numpy.zeros((sizex,sizey)))
norm[x,y]=4
if x > 0:
norm[x-1,y]=-1
if x < sizex-1:
norm[x+1,y]=-1
if y > 0:
norm[x,y-1]=-1
if y < sizey-1:
norm[x,y+1]=-1
S[x*sizex+y,:] = norm.flatten()
S=numpy.mat(S)
return S*S.T