Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: A "New" Idea for Adaptive Programs

Author: Rafael Andrist

Date: 14:04:34 12/23/02

Go up one level in this thread


>Perhaps a more practical approach might be to have a single program with many
>parameters that could be set or reset quickly.  In this scenario, the stimuli
>would result in calling of a "reconfigure" program.  This program would then,
>very quickly, reset the parameters of the main program.  In this way, the main
>program would, in effect, be a new program.
>
>A more generalized version of the latter approach would be for the "reconfigure"
>program to VERY QUICKLY make extensive modifications to the main program.
>
>Are these ideas any good?
>
>If they are, are they currently in use in existing chess engines?

If you "only" talk about changing parameters of the evaluation function, there
is program called KnightCap which seems to do so by an algorithm named
temporal-difference learning. The source code should be available on the net,
however I was unable to access it :-(

http://www.syseng.anu.edu.au/lsg/knightcap.html

regards
Rafael B. Andrist



This page took 0 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.