Which implementation are you looking for?

You can use genetic algorithms in ANNs for different purposes:

1) Weight adjustment

2)Architecture generation (e.g. select the number of nodes of the hidden layer)

3) Generate both architecture and weights

thank you a lot.

I want to use GA to optimize the achitecture of ANN in order to increase the accuracy of forcasting . so i think that include the weights and number of nodes of the hidden layer.

I would suggest that you look at

Evolving Artificial Neural Networks by Xin Yao(available online). The reference is kind of dated but gives a very good introduction to the subject(and from there expand to more current literature).

After that based on the specific problem, you need to select what parts of the architecture do you want to encode in the chromosome and how to encode the weights(binary or real valued).

Should you need implementation details you can post additional question or directly write to me and I will be glad to assist in the process with implementation details.

Alternatively, you may wish to use a off-the-self software to do the job for you. I suggest using NeuroSoulutions. This program lets you use GA to optimize, number of nodes of the hidden layer, learning rate and momentum coefficient without any programming!. All you need to do is setting a few parameters like population size, max generation. Finally, GA operators such as selection, crossover and mutation and their probabilities can also be set from the "Advanced Genetic Options" menu too. I use NeuroSoulutions ver. 5.07.

Thank you very much for your answers.