Author: Graham Laight
Date: 03:14:38 10/13/00
Go up one level in this thread
On October 12, 2000 at 12:41:54, Dan Andersson wrote: >One could go the whole nine yards in generalizing by applying Genetic >Programming to the search algorithm selection and get a program with almost no a >priori knowledge. What a good idea! A self learning program that didn't require a human to add knowledge bit by bit! I'm not entirely clear how it would work, though. Can anyone explain it in a simple, straightforward way? In particular, what, exactly, would the genetic algorithm (GA) be "evolving"? Here's my guess at what I think Dan meant: have a pile of evaluation components, and use a GA to combine them in a way that effectively selects the best lines to search. If this is correct, you've still got the problem that different types of position would have different types of move selection criteria. Maybe you could have more genetic algorithms at a higher level determining different position types? This is VERY exciting - we could be on our way to designing a practical, implementable, self learning chess machine here! -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.