NeuralNetwork_matlab/MLP_sequential.m at main · GSSfearless/NeuralNetwork_matlab · GitHub

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
function [acc_train,acc_test] = MLP_sequential(input_train,input_test,n,epoch,lr,rgl,Fcn)
[m1,n1] = size(input_train);
[m2,n2] = size(input_test);

train_data = input_train(1:m1-1,:);
train_label = input_train(m1,:);
test_data = input_test(1:m2-1,:);
test_label = input_test(m2,:);

test_data = repmat(test_data,1,9);
test_label = repmat(test_label,1,9);

train_data = num2cell(train_data,1);
train_label = num2cell(train_label,1);
test_data = num2cell(test_data,1);
test_label = num2cell(test_label,1);

net = patternnet(n);
net.trainParam.epochs = epoch;
net.trainParam.lr = lr;
net.performParam.regularization = rgl;
net.trainFcn = Fcn;
net.performFcn = 'crossentropy';
net.trainParam.showWindow = false;
net = configure(net,train_data,train_label);

train_result = round(cell2mat(net(train_data)));
acc_train = 1 - abs(mean(train_result - ...
    cell2mat(train_label)));

% test_data = cell2mat(test_data);
% test_label = cell2mat(test_label);
% acc_test = mean(1 - abs(net(test_data) - test_label));
test_result = round(cell2mat(net(test_data)));
acc_test = 1 - abs(mean(test_result - ...
    cell2mat(test_label)));

% display(['accuracy of training data:',acc_train]);
% display(['accuracy of testing data:',acc_test]);