-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathfirst.m
248 lines (215 loc) · 5.74 KB
/
first.m
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
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
start=time();
fid = fopen('train-images.idx3-ubyte', 'r', 'b');
header = fread(fid, 1, 'int32')
totalimages = fread(fid, 1, 'int32')
numrows = fread(fid, 1, 'int32')
numcols = fread(fid, 1, 'int32')
img=ones(numrows,numcols,totalimages);
for k=1:totalimages
%disp(k);
img(:,:,k)= fread(fid, [numrows,numcols], 'uchar');
endfor
fclose(fid);
%for k=1:totalimages
% disp("image:");
% for i=1:numrows
% for j=1:numcols
% printf("%d ",img(k,i,j));
% endfor
% printf("\n");
% endfor
% if k==2
% break;
% endif
%endfor
%disp(totalimages);
fid2 = fopen('train-labels.idx1-ubyte', 'r', 'b');
header1 = fread(fid2, 1, 'int32')
totallabels = fread(fid2, 1, 'int32')
labels= fread(fid2, totallabels, 'uchar');
fclose(fid2);
disp('input readed');
disp(time()-start);
%for i=1:totallabels
% header1 = fread(fid2, 1, 'uchar');
% labels(i)=header1;
%endfor
hiddenlayers=90;
wji=rand(numcols*numrows,hiddenlayers)*0.09; %num of hidden nodes is 15 bias is 0th row
wkj=rand(hiddenlayers,10)*.09;
eta=0.5;
threshold=1;
it=0;
J=102;
format long
while(J>100)
J=0;
for k=1:totalimages
xi=[];
for i=1:numrows
xi=[xi img(i,:,k)]; %xi=kth row from inage matrix,convert xi to 784*1
endfor
xi=double(xi)/255;
t(1:10,1)=0.02;
tk=labels(k);
t(tk+1,1)=0.95; %index starts from 1 thats why tk+1
netj=(xi*wji)'; %wj0=0,bias=0,net input on the node of hidden layer dimension-no of hidden nodes, wji ka transpose and then multiply it with xi => 15*1
yj=1./(1+(exp(-netj)));
netk=(yj'*wkj)'; %wk0=0,bias=0,net input on the node of output layer dimension-no of output nodes,10*1
zk=1./(1+(exp(-netk))); %10*1
dfk=zk.*(1-zk); %10*1
dfj=yj.*(1-yj); %15*1
dk=(t-zk).*dfk; %10*1
dwkj=eta*yj*dk';
wkj=wkj+dwkj;
dj=dfj.*(wkj*dk); %summation aayega! 10*1
dwji=eta*xi'*dj';
wji=wji+dwji;
%wkj=wkj+dwkj- (0.02*eta*wkj); %weight Decay!
%wji=wji+dwji-(0.02*eta*wji);
J=J+1/2*sum((t-zk).^2);
% J=J+1/2*sum((t-zk).^2);
endfor
%disp(it);
disp(J);
if mod(it,1000) == 0
disp(it);
disp(J);
dlmwrite('ipweights.txt',wji);
dlmwrite('hiddenweights.txt',wkj);
endif
it=it+1;
endwhile
dlmwrite('inputweights90.txt',wji);
dlmwrite('hiddenweights90.txt',wkj);
disp(it);
disp(time()-start);
validationimg=totalimages/5;
testimg=ones(numrows,numcols,validationimg);
for i=1:validationimg
testlabels(i,:)=labels(i,:);
testimg(:,:,i)=img(:,:,i);
endfor
w22=size(testlabels);
w222=size(testimg);
trainimg=ones(numrows,numcols,totalimages-validationimg);
for i=validationimg+1:totalimages-validationimg
trainlabels(i,:)=labels(i,:);
trainimg(:,:,i)=img(:,:,i);
endfor
answer=zeros;
for j=1:5
[confusionmat accuracy]=fivefoldval(trainimg,testimg,trainlabels,testlabels,validationimg);
answer=answer+confusionmat;
endfor
per=accuracy/(validationimg*5);
printf("First fold\n");
printf("Accuaracy: %f\n",per*100-1);
printf("Error Rate: %f\n",(1-per)*100);
printf("Precision: %f\n",per);
printf("Recall: %f\n",per);
disp(confusionmat);
%2nd
for i=1+validationimg:2*validationimg
testlabels(i,:)=labels(i,:);
testimg(:,:,i)=img(:,:,i);
endfor
w22=size(testlabels);
w222=size(testimg);
trainimg=ones(numrows,numcols,totalimages-validationimg);
for i=1:validationimg
trainlabels(i,:)=labels(i,:);
trainiimg(:,:,i)=img(:,:,i);
endfor
for i=2*validationimg+1:totalimages
trainlabels(i,:)=labels(i,:);
trainiimg(:,:,i)=img(:,:,i);
endfor
answer=zeros;
for j=1:5
[confusionmat accuracy]=fivefoldval(trainimg,testimg,trainlabels,testlabels,validationimg);
answer=answer+confusionmat;
endfor
per=accuracy/(validationimg*5);
printf("Second fold\n");
printf("Accuaracy: %f\n",per*100-1);
printf("Error Rate: %f\n",(1-per)*100);
printf("Precision: %f\n",per);
printf("Recall: %f\n",per);
disp(confusionmat);
%3rd
for i=validationimg*2+1:3*validationimg
testlabels(i,:)=labels(i,:);
testimg(:,:,i)=img(:,:,i);
endfor
w22=size(testlabels);
w222=size(testimg);
trainimg=ones(numrows,numcols,totalimages-validationimg);
for i=1:2*validationimg
trainlabels(i,:)=labels(i,:);
trainiimg(:,:,i)=img(:,:,i);
endfor
for i=3*validationimg+1:totalimages
trainlabels(i,:)=labels(i,:);
trainiimg(:,:,i)=img(:,:,i);
endfor
answer=zeros;
for j=1:5
[confusionmat accuracy]=fivefoldval(trainimg,testimg,trainlabels,testlabels,validationimg);
answer=answer+confusionmat;
endfor
per=accuracy/(validationimg*5);
printf("Third fold\n");
printf("Accuaracy: %f\n",per*100-1);
printf("Error Rate: %f\n",(1-per)*100);
printf("Precision: %f\n",per);
printf("Recall: %f\n",per);
disp(confusionmat);
%4th
for i=validationimg*3+1:4*validationimg
testlabels(i,:)=labels(i,:);
testimg(:,:,i)=img(:,:,i);
endfor
trainimg=ones(numrows,numcols,totalimages-validationimg);
for i=1:3*validationimg
trainlabels(i,:)=labels(i,:);
trainiimg(:,:,i)=img(:,:,i);
endfor
for i=4*validationimg+1:totalimages
trainlabels(i,:)=labels(i,:);
trainiimg(:,:,i)=img(:,:,i);
endfor
answer=zeros;
for j=1:5
[confusionmat accuracy]=fivefoldval(trainimg,testimg,trainlabels,testlabels,validationimg);
answer=answer+confusionmat;
endfor
per=accuracy/(validationimg*5);
printf("Fourth fold\n");
printf("Accuaracy: %f\n",per*100-1);
printf("Error Rate: %f\n",(1-per)*100);
printf("Precision: %f\n",per);
printf("Recall: %f\n",per);
disp(confusionmat);
%5th
for i=validationimg*4+1:5*validationimg
testlabels(i,:)=labels(i,:);
testimg(:,:,i)=img(:,:,i);
endfor
trainimg=ones(numrows,numcols,totalimages-validationimg);
for i=1:4*validationimg
trainlabels(i,:)=labels(i,:);
trainiimg(:,:,i)=img(:,:,i);
endfor
answer=zeros;
for j=1:5
[confusionmat accuracy]=fivefoldval(trainimg,testimg,trainlabels,testlabels,validationimg);
answer=answer+confusionmat;
endfor
per=accuracy/(validationimg*5);
printf("Fifth fold\n");
printf("Accuaracy: %f\n",per*100-1);
printf("Error Rate: %f\n",(1-per)*100);
printf("Precision: %f\n",per);
printf("Recall: %f\n",per);
disp(confusionmat);