Author: Vincent Diepeveen
Date: 05:02:42 12/21/99
Go up one level in this thread
On December 20, 1999 at 15:29:06, James Swafford wrote: >Hi Tijs, > >I think we're in the same boat. I am setting up a genetic >algorithm experiment, hopefully to begin late next summer. For some reason i see a clear difference between 'genetic' and 'neural learned' parameters. The big problem is to find a good and objective test for a newly created parameterset. If the genetic algorithm just flips a single bit somewhere in the paramater set, then it's *very* hard to determine whether it's a good bitflip or a wrong one. >I've resigned myself to the fact that there won't be any >quick results. In fact, I'm expecting this to take at least >a couple years to derive anything meaningful. >TDChess (KnightCap) used neural networks. You can find >a link to KnightCap on my site... > >http://members.xoom.com/jswaff/chessprg > >Good luck.. please keep up posted! > >-- >James > > >On December 20, 1999 at 15:00:50, Tijs van Dam wrote: > >>Hello, >> >>For my simple chess program GKJunior I now stand for the tremendous task of >>writing a decent evaluation function. I took a look at the 3000 lines of >>crafty's eval and decided that i just don't know that much of chess itself to >>ever do this. >> >>Therefore, I decided i would like to have the program do some stuff by itself. >>Sounds easy, but isn't. I'm looking in the direction of neural programming... I >>want the program to calculate a value for positional advantage, knowing nothing >>at all of passed pawns, trapped bishops, connected rooks or whatsoever. >> >>Has anybody here some experience on this topic and could you point me in some >>direction (books, internet resources)? >> >>Greetings, >>Tijs van Dam
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.