I just wanted to know if most of you use a GA to select weights for your NN? I''m trying this method for a NN that plays mancala for a programming compo, and it seems to work quite well. How complicated is calculating the weights based on desired outputs for specific inputs? I''m new to AI, so any details or advice would be helpful.

Backpropogation training isn''t particularly complex, although many people seem to shy away from it because of the mathematics. There are plenty of resources online that explain the method fairly clearly. I''d suggest you take a look and compare the quality of results and training times to your current method, then choose which one you like the best.

Cheers,

Timkin

theres actually a course in a.i i took last year at university and the lecturer notes are online and explain these methods quite clearly and there is also sample code on how to code up something like this, its pretty high level and doesnt get nitty gritty on the maths.all sample code are in java, notes in ps and pdf

http://www.cs.ucc.ie/~dgb/courses/ai/lectures.html