Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: knowing when you've improved your evaluation function

Author: Anthony Cozzie

Date: 11:55:23 04/15/04

Go up one level in this thread


On April 15, 2004 at 14:51:26, Anthony Cozzie wrote:

>On April 14, 2004 at 04:48:50, Dave Gomboc wrote:
>
>>You may be interested in checking out my M.Sc. Thesis, "Tuning Evaluation
>>Functions by Maximizing Concordance".  An unofficial version of it is now
>>available at http://www.cs.ualberta.ca/~dave.
>>
>>The purpose of the work is to provide a metric by which one can assess whether a
>>change to their heuristic evaluation function has been beneficial or not.  It
>>includes and expands upon work reported at ACG-10 in Graz.
>>
>>Feel free to post comments / questions / criticisms also!  If there's something
>>I can quickly address (e.g. typos) I can perhaps correct it before I hand it in.
>>
>>If you have or know someone who has back issues of Computer-Schach und Spiele,
>>could you please check if I have the proper references for the Nunn tests (see
>>Nunn 1998 and Nunn 2000 in the reference section).  Earlier Fred Friedel put me
>>into contact with Dieter Steinwender, but I haven't heard from him in a week or
>>so now, and I want to be 100% sure I have it right.
>>
>>Also, readers who know Dutch are welcome to verify that I've done the proper
>>thing with Martin van der Meulen.
>>
>>Dave
>
>As I understand, your paper goes like this (omitting background, proofs,
>implementation details, results, little things like that :)
>
>1. Get 650,000 positions from Chess Informant with GM evals (e.g. +-, =, etc).
>
>2. Divide positions into sets of +-, +=, =, etc.
>
>3. Tune eval such that all the positions in the set '+-' evaluate higher than
>the positions in the set '+=' which all evaluate higher than positions in the
>set '=' (etc) using gradient search.
>
>Close?
>
>anthony

Its actually a pretty interesting paper, because it gets around the problem of
getting a numerical evaluation out of GM.

Criticism: In Acrobat, your pages are offset.  They look like:

       PPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPP
   PPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPP
       PPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPPP

anthony



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.