forked from nachiram/RVFL-matlab
-
Notifications
You must be signed in to change notification settings - Fork 0
/
demo_abalone.m
150 lines (119 loc) · 4.04 KB
/
demo_abalone.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
clear
clc
abalone_R;
data=data(:,2:end);
dataX=data(:,1:end-1);
% do normalization for each feature
mean_X=mean(dataX,1);
dataX=dataX-repmat(mean_X,size(dataX,1),1);
norm_X=sum(dataX.^2,1);
norm_X=sqrt(norm_X);
norm_X=repmat(norm_X,size(dataX,1),1);
dataX=dataX./norm_X;
dataY=data(:,end);
abalone_conxuntos;% for datasets where training-testing partition is available, paramter tuning is based on this file.
trainX=dataX(index1,:);
trainY=dataY(index1,:);
testX=dataX(index2,:);
testY=dataY(index2,:);
MAX_acc=zeros(4,1);
Best_N=zeros(4,1);
Best_C=zeros(4,1);
Best_S=zeros(4,1);
S=-5:0.5:5;
for s=1:numel(S)
for N=3:20:203
for C=-5:14
Scale=2^S(s);
option1.N=N;
option1.C=2^C;
option1.Scale=Scale;
option1.Scalemode=3;
option1.bias=0;
option1.link=0;
option2.N=N;
option2.C=2^C;
option2.Scale=Scale;
option2.Scalemode=3;
option2.bias=1;
option2.link=0;
option3.N=N;
option3.C=2^C;
option3.Scale=Scale;
option3.Scalemode=3;
option3.bias=0;
option3.link=1;
option4.N=N;
option4.C=2^C;
option4.Scale=Scale;
option4.Scalemode=3;
option4.bias=1;
option4.link=1;
[train_accuracy1,test_accuracy1]=RVFL_train_val(trainX,trainY,testX,testY,option1);
[train_accuracy2,test_accuracy2]=RVFL_train_val(trainX,trainY,testX,testY,option2);
[train_accuracy3,test_accuracy3]=RVFL_train_val(trainX,trainY,testX,testY,option3);
[train_accuracy4,test_accuracy4]=RVFL_train_val(trainX,trainY,testX,testY,option4);
if test_accuracy1>MAX_acc(1); % paramater tuning: we prefer the parameter which lead to better accuracy on the test data.
MAX_acc(1)=test_accuracy1;
Best_N(1)=N;
Best_C(1)=C;
Best_S(1)=Scale;
end
if test_accuracy2>MAX_acc(2);% paramater tuning: we prefer the parameter which lead to better accuracy on the test data.
MAX_acc(2)=test_accuracy2;
Best_N(2)=N;
Best_C(2)=C;
Best_S(2)=Scale;
end
if test_accuracy3>MAX_acc(3);% paramater tuning: we prefer the parameter which lead to better accuracy on the test data.
MAX_acc(3)=test_accuracy3;
Best_N(3)=N;
Best_C(3)=C;
Best_S(3)=Scale;
end
if test_accuracy4>MAX_acc(4);% paramater tuning: we prefer the parameter which lead to better accuracy on the test data.
MAX_acc(4)=test_accuracy4;
Best_N(4)=N;
Best_C(4)=C;
Best_S(4)=Scale;
end
end
end
end
abalone_conxuntos_kfold; %for datasets where training-testing partition is not available, performance vealuation is based on cross-validation.
ACC_CV=zeros(4,1);
for i=1:4
trainX=dataX(index{2*i-1},:);
trainY=dataY(index{2*i-1},:);
testX=dataX(index{2*i},:);
testY=dataY(index{2*i},:);
option1.N=Best_N(1);
option1.C=2^Best_C(1);
option1.Scale=Best_S(1);
option1.Scalemode=3;
option1.bias=0;
option1.link=0;
option2.N=Best_N(2);
option2.C=2^Best_C(2);
option2.Scale=Best_S(2);
option2.Scalemode=3;
option2.bias=1;
option2.link=0;
option3.N=Best_N(3);
option3.C=2^Best_C(3);
option3.Scale=Best_S(3);
option3.Scalemode=3;
option3.bias=0;
option3.link=1;
option4.N=Best_N(4);
option4.C=2^Best_C(4);
option4.Scale=Best_S(4);
option4.Scalemode=3;
option4.bias=1;
option4.link=1;
[train_accuracy1,ACC_CV(1,i)]=RVFL_train_val(trainX,trainY,testX,testY,option1) ;% ACC_CV each row is the accuracy for one RVFL configuration. Each column is a single trial for evaluation.
[train_accuracy2,ACC_CV(2,i)]=RVFL_train_val(trainX,trainY,testX,testY,option2);
[train_accuracy3,ACC_CV(3,i)]=RVFL_train_val(trainX,trainY,testX,testY,option3);
[train_accuracy4,ACC_CV(4,i)]=RVFL_train_val(trainX,trainY,testX,testY,option4);
end
mean(ACC_CV,2)