Hey,

For a simple network, solving for a quadratic function every time after training I get a really varied number of epochs can any one tell me why?!?! It doesnt seem to be very consistent

And also, does anyone know some code or WebPages which refers to automatically checking if the performance goal has been reached ?

CODE:

y = (x).^2 -x + 5;

P = [-5, 0, 2, 5, 9, 12, 15, 20, 22, 35, 41];
T = [35,5,7,25,77,137,215,385,467,1195,1645];

%Creating the neural network - with feed forward backprog. 2hidden layers 1 output layer, tansig transfer function

net = newff([a], [8, 12, 1], {'tansig' 'tansig' 'purelin'}, 'trainlm', 'learngdm', 'mse');

net = init(net);
net.trainParam.show = 50;
net.trainParam.epochs = 5000;
net.trainParam.goal = .5;

[net] = train(net, P, T); %training the network

PPP = [15, 23, 31, 9]

for i = 1
TTT = sim(net, PPP);
NN(i, :) = TTT;
end

%Displaying the results
disp( 'Output')
disp (NN)

Thanks,

Rach