From: Optimization of underwater wet welding process parameters using neural network
Program algorithm | |
---|---|
load matlab.mat | |
% inputs | |
I=DataProject(:,1); | D=DataProject(:,4); |
U=DataProject(:,2); | H=DataProject(:,5); |
v=DataProject(:,3); | F=[I U v D H]; |
%outputs | |
W=DataProject(:,6); | Â |
P=DataProject(:,7); | G=[W P R]; |
R=DataProject(:,8); | Â |
% training | Â |
p=F(1:12,:); | t=G(1:12,:); |
% testing | Â |
x=F(13:16,:); | Z=[x y]; |
y=G(13:16,:); | Â |
% form the network | Â |
net=feedforwardnet([40],'trainscg'); | net.trainParam.max_fail=2000; |
net.trainParam.goal=0; % error goal | net.trainParam.lr=0.001; |
net.trainParam.epochs=3000; % maximum iterations | net.trainParam.mc=0.9; |
net.trainParam.show=25; % showing intervals | Â |
% Network initialization | Â |
net.initFcn='initlay'; | [net,tr]=train(net,p',t'); % training the net |
net.layers{1}.initFcn='initnw'; | view(net) |
net=init(net);% initialize the net (weights and biases initialized) | Â |
% simulating the network with training inputs for testing | Â |
f=net(x'); | f' |
% compare results/target | Â |
Error=f'-y | Â |