Thread Subject: Neural Networks and Genetic Algorithm

Hi All of You
How to optimize Feed Forward Neural Network output using Genetic Algorithm instead of backpropagation Algorithm. i.e. to used GA for weights optimization.
A small Matlab code will help me alot
thanks

"zaheer ahmad" <ahmad.zaheer@yah00000.com> wrote in message <gmv4l2$7ib$1@fred.mathworks.com>...
> Hi All of You
> How to optimize Feed Forward Neural Network output using Genetic Algorithm instead of backpropagation Algorithm. i.e. to used GA for weights optimization.
> A small Matlab code will help me alot
> thanks

That should be simple.

If you're dealing with a backprop NN with 1 hidden layer consisting 3 hidden nodes, and 2 input, then you have some (2x3)+(3x1) = 9 connection weights. Using gatool, you can specify in the Problem section of the GUI, that you have 9 variables. You can lower and/or upper bounds of these weights as you wish.

And then, gatool will do the rest.

Good luck!

On Mar 4, 6:59=A0am, "Akmal Aulia" <akmalt...@yahoo.com> wrote:
> "zaheer ahmad" <ahmad.zah...@yah00000.com> wrote in message <gmv4l2$7i...=
@fred.mathworks.com>...
> > Hi All of You
> > How to optimize Feed ForwardNeuralNetwork output using Genetic Algorith=
m instead of backpropagation Algorithm. i.e. to used GA for weights optimiz=
ation.
> > A small Matlab code will help me alot
> > thanks
>
> That should be simple.
>
> If you're dealing with a backprop NN with 1 hidden layer consisting 3 hid=
den nodes, and 2 input, then you have some (2x3)+(3x1) =3D 9 connection wei=
ghts. =A0Using gatool, you can specify in the Problem section of the GUI, t=
hat you have 9 variables. =A0You can lower and/or upper bounds of these wei=
ghts as you wish.
>
> And then, gatool will do the rest.

The number of weights is

Nw =3D (2+1)*3 + (3+1)*1 =3D 13

Hope this helps.

Greg