Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Pattern Recognition and Chess

Author: Michael Neish

Date: 18:04:15 10/11/00

Go up one level in this thread



Thanks to the three of you who replied to my question.  I knew about using GA's
to tune the evaluation parameters, but I was mainly wondering whether they (and
NN's) had been applied more directly, i.e., to recognising patterns on the board
itself.

About human NPS, I think it's very difficult to estimate such a value, since
most players have their own mental database which recognises patterns and knows
intuitively what they mean.  In this way they can make fast decisions without
really calculating anything, which probably counts as a few tens of thousands of
nodes, or more.

Well fast searching is certainly winning at the moment.  Who knows about the
future.  At the risk of being stamped on quite heavily, I don't think it counts
as AI, since it's just a higgledy-piggledy way of throwing a mass of empirical
rules of thumb together in a way that seems to play good Chess.  I just thought
it might be fun to write a program that uses a different technique, even if its
strength initially goes down about a thousand points!  It might never make it up
again, but at least one will have learned a lot doing it.

Have any of you tried a different approach to the problem?

Cheers,

Mike.




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.