%% Sample Training Session

%Create some artificial data
P=linspace(-2,2,50);
T=sin(pi*P)+0.2*randn(size(P));

% We'll build a 1-15-1 network and train it:
net=feedforwardnet(5); %10 nodes in hidden layer
net=configure(net,P,T); % Initialize weights and biases
y1=sim(net,P);          % Just for fun (randomly assigned weights)

% Training:
net=train(net,P,T);     %Training command
y2=sim(net,P);          % Network output after training

%% See how the network performs with new data
tt=linspace(-2,2,200);
yy=sim(net,tt);

%% Compare to the noise-free model
yModel=sin(pi*tt);

plot(P,T,'o',P,y1,'b-',tt,yy,'k-',tt,yModel,'r-');
legend('Data','Output 1','Output 2','No Noise');