Author: Dann Corbit
Date: 09:49:59 01/27/99
Go up one level in this thread
On January 27, 1999 at 05:08:59, Frank Schubert wrote: >Dear computerchess friends and program authors, > >last year the german academy for chess and sciences was founded at Karpov chess >centre in Baden-Baden (D). As a co-ordinator for the topic 'Chess and Neural >Networks' I just try to get an overall picture of interesting activities in this >field. At the moment I only know the program ALEXS by van Tiggelen (NL) and the >NeuroChess-project of Sebastian Thrun (Uni Pittsburgh). Who can give me more >informations about chess (programs) and Neural Networks ? > >Additionally it would be nice if we can start a discussion here on CCC about >possible applications of Neural Networks in chess programs. So does someone of >the program authors plan to use NN's in his program ? What is your opinion about >it ? What are your experiences in this field and what will be the future trends >? Any contributions to this subject are welcome. Neural net programs do very poorly. I think it is because they have no temporal insight (e.g. they see it is good to develop the queen, so they thrust it out right away). I think a very good application of NN to chess would be to simplify the decision making process needed. For instance, give the NN only a few simple parameters: This move has win/loss/draw value x. It has computer eval of z. It has GM player choice of q. That move has win/loss/draw value a. It has computer eval of b. It has GM player choice of c. Which move should I choose? For a simple application such as this, I think they may have great value. But right now, it is asking way too much of them to play the entire game. If we model the way a brain works, but make it a teeny-tiny pipsqueak of a brain, where is the benefit? Whatever we do with it, we better not overtax it. We might give it a migraine.
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.