bp2.m

this code for backpropagation on matlab. with indonesian language

% backpropagation dengan gradien descend(learngd)
clear;

% data input dan target
P=[0 1 2 1 10 12 -5 -8 -10 -15; 0 1 -1 6 3 -1 -2 2 -5 2];
T=[0 0 1 1 2 2 -1 -1 -2 -2];

% membangun jaringan feedforward
net=newff(minmax(P), [5 1], {'tansig' 'purelin'});

%melihat bobot awal input lapisan, dan bias
bobotawal_input         = net.IW{1,1}
bobotawal_bias_input    = net.b{1,1}
bobotawal_lapisan       = net.LW{2,1}
bobotawal_bias_lapisan  = net.b{2,1}

% set fungsi pelatihan jaringan
net.adaptFcn = 'adaptwb';

%set fungsi pembelajaran bobot input , lapidan dan bias
net.inputWeights{1,1}.learnFcn='learngd';
net.layerWeights{2,1}.learnFcn='learngd';
net.biases{1,1}.learnFcn='learngd';
net.biases{2,1}.learnFcn='learngd';

% set learning rate 
net.inputWeights{1,1}.learnParam.lr=0.02;
net.layerWeights{2,1}.learnParam.lr=0.02;
net.biases{1,1}.learnParam.lr=0.02;
net.biases{2,1}.learnParam.lr=0.02;

% set maksimum epoch
net.adaptParam.passes = 50;

%ubah matrik input dan target ke bentk cell
p=num2cell(P,1);
t=num2cell(T,1);

%melakukan pelatihan
[net,Y,e]=adapt(net,p,t);

% melihat bobot awal input, lapisan dan bias
bobotawal_input         = net.IW{1,1}
bobotawal_bias_input    = net.b{1,1}
bobotawal_lapisan       = net.LW{2,1}
bobotawal_bias_lapisan  = net.b{2,1}

% melakukan simulasi
a = sim(net,P)

% menggambar grafik
subplot(211)
plot(P(1,:),T,'bo',P(1,:),a,'r*');
title('Perbandingan antara target (o) dan output jaringan (*)');
xlabel('input pertama');
ylabel('target atau ooutput');
grid;
subplot(212)
plot(P(2,:),T,'bo',P(2,:),a,'r*');
title('Perbandingan antara target (o) dan output jaringan (*)');
xlabel('input kedua');
ylabel('target atau ooutput');
grid;