Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Chess, Backgammon and Neural Nets (NN)

Author: Peter Fendrich

Date: 11:38:41 08/20/98

Go up one level in this thread


On August 20, 1998 at 12:23:33, Torsten Schoop wrote:

>Hi all,
>
>does anyone know why NN backgammon programs play very well (TD-Gammon,
>JellyFish, Snowie)but the NN chess playing programs do not?
>
>Ciao
>Torsten

In my opinion it's all about the "degree of freedom". I hope this expression
make sence in english...
From my experience at least the classical backpropagation nets are quite
unsensitive to this up to a certain point. After that the net performance drops
considerably and it doesn't help to extend the number of nodes or to insert more
layers even if you have all the training material in the world. Finally, I think
it's much easier to find a well covering set of training patterns in Backgammon
compared to chess. I can't see how it's possible to represent the chess domain
in a net, covering all or at least most of the different aspects of chess
positions that there is. All chess rules combined with tactical issues combined
with positional issues in different phases of the game is too much!

Another answear could be that they just haven't found the right net types and
topologies, yet... :)
//Peter



This page took 0.01 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.