Computer Chess Club Archives




Subject: Re: Zappa Report

Author: Ingo Althofer

Date: 04:01:46 12/31/05

Go up one level in this thread

On December 31, 2005 at 04:38:46, Rafael Andrist wrote:
>On December 30, 2005 at 13:02:43, Ingo Althofer wrote:
>>... in case of linear evaluation functions with lots
>>of terms there is always a small subset of the terms
>>such that this set with the right parameters is
>>almost as good as the full evaluation function.
> Can you state this a bit more precise, please?
> As long as it is not clear what you mean by "lots of" and "small"
> and "almost as good", your statement says nothing and could be
> trivially true, if, for ex.
> "small" just means "less or equal".

I can assure you that the result is not trivial.

Unfortunately I wrote the paper, before the
mathematical community developed fully "electronized"
publication services. So you have to look at the
paper versions:

The technical result with complete proofs is given in
* I. Althofer. On sparse approximations to randomized strategies and convex
combinations. Linear Algebra and Applications 199 (1994), 339-355.

The application to linear evaluation functions in chess, go,
and other games is described in
* I. Althofer. On telescoping linear evaluation functions.
ICCA-J , Vol 16 (1993), pp. 91-94 (issue June 1993).

Funnily, in this context "Go" seems to be only about
2-3 times more complicated than chess, if I remember correctly.
(... where complexity is measured in number of linear terms needed to
achieve a certain level of quality)

Regards, Ingo Althofer.

This page took 0.07 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.