Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Philosophical thoughts: Probability Eval vs. Conventional Evaluation

Author: Wylie Garvin

Date: 22:04:42 01/19/02

Go up one level in this thread


Hi,

  I just wanted to mention some research I came across on a topic close to this.
You may find it interesting.  The approach proposed in this research is called
Partial Order Bounding.

http://www.cs.ualberta.ca/~mmueller/search/pob.html

Here is a brief excerpt from the paper:
-----
This paper describes a different, more radical approach to the basic questions
of "when" and "how" to compare evaluations.  Since many pairs of positions are
really incomparable from the limited knowledge and accuracy incorporated in an
evaluation function, it seems that forcing a comparision by mapping all
positions to the same numeric scale does more harm than good.  From this point
of view, using only a partial ordering of positions seems to be a natural
approach.  The main technical contribution of this paper is an efficient method
for combining partial order evaluation with minimax search.  Partial order
bounding (POB) provides a way to use partial order position evaluation in
conjunction with any minimax-based search engine such as alpha-beta or
proof-number search.
-----

wylie



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.