Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: How to Implement Pattern Recognition

Author: Graham Laight

Date: 02:45:51 11/19/00

Go up one level in this thread


On November 18, 2000 at 18:40:55, Fernando Villegas wrote:

>Hi :
>I suppose that what you ask supposes an amount of memory at least so big as a
>the brain has. Human brain have around 14 billions neurons into which patterns
>are stored. But what is more, each neuron can asociate many times with other, so
>in fact the real number of units of memory is far greater, perhaps 100 times. Of
>course for doing just chess task you does not need so much, but even so the
>amount is GREAT. Anyway I cannot see how a electronic cell that is just 0 or 1
>can at the same time "be" another thing to another cell. Maybe the answer goes
>for a more simple way than yours: not to store patterns, but to stores isolated
>features of patterns, so a total pattern is no more than a tablebase. In other
>words, the program perhaps does not need the pattern as such, as an specific
>object to store, but its components, that we and the program knows are part of
>it.
>Fernando

I'm not "advocating" this idea yet, but I think it's worthy of further
discussion.

I think that something like a tablebase to store patterns is a good idea. Then,
for each pattern, a data mining investigation into a chess games database might
be possible, to estimate the likely contribution of this pattern to defeat or
victory. Then maybe an evaluation could be made from the combination of
patterns.

Question - how would useful patterns in the existing games be recognised?  Also,
might there be useful patterns in there that humans haven't recognised yet?

It all sounds fanciful at the moment, and I'm have a feeling one would run into
a lot of problems trying to implement it. However, it has the vitue of trying to
tease out the knowledge that's stored in chess databases. Perhaps we could think
about that some more?

-g



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.