I still havent been able to crack my ANN problem. Instead of posting source and ask for help, I ask if anyone could tell me if I got the maths right or not. First off, heres how I describe different parts of the net: w(l,f,n) - Weight for input f to neuron n in layer l. x(l,n) - Output from neuron n in layer l. l=0 is inputs from envoirment and n=0 is the bias. d(l,n) - Deltas for neuron n in layer l. Used in the back propagation function. Now, heres the layout of my net: I calculate the delta for the output neuron with this forula: // output layer delta, we only have one output // d(2,0) = x(3,0)(1 - x(3,0))(d - x(3,0)) Deltas for the hidden layers are calculated like this: // d(l,n) = x(l,n) (1 - x(l,n)) w(l+1,n,n-1) d(l+1,n) Now I alter the weights, thats done with this formula (h is learning constant): // Deltas calculated, now alter the weights // w(l,f,n) = h * x(l,f) * d(l,n) I have gone through the code several times, I cannot find anything wrong. So Ive made the conclusion that I must have missunderstood the maths. Could anyone tell me if these formulas would work in a neural net? If the math is correct, at least I know the error is somewhere in the code. Anyone, please^2 help me out with the math part. Thanks in advance. [smile]