Author: Stefano Gemma
Date: 23:31:39 04/28/04
Go up one level in this thread
On April 28, 2004 at 17:44:49, Pat King wrote: >On April 27, 2004 at 17:03:25, Roger D Davis wrote: > >>On April 27, 2004 at 16:33:56, Pat King wrote: >> >>>I've tried genetic algorithms before. Depending on how you implement them, >>>convergence will be extremely slow to non-existent! >>> >>>you just >>>end up with a bunch of more or less random weights. >>> >>>Pat King >> >>Seems like you'd want to do some tuning using situations where one move is known >>to be better. Then you can tune for more subtle situations. >> >>Roger > >The whole point of GA, IMHO, is that the tuning process can be pretty much >"hands off". If you've a group of test positions for which the best moves are >known, there are many other techniques that will give quicker and better >results, at least in terms of matching the training set. > >OTOH, you could be arguing as I (meant to have) said earlier: GA doesn't work >well at all for chess. In that case, we are totally in agreement :) If this is true... then i'm wasting my time ;-). But, as you know, there are not just "one" GA. As said in the other post, you can think "genetically" for the tactic part or for the strategic part of the program. The tactic maybe is better done by alfa-beta and its variations. The strategic should be done from something using AI. A good evaluation function is nothing more (IMHO!) than an "hard coded" expert system. As you surely know, ES works well when you can exactly define the knowledge that you need. This is true for chess, because you can talk about some euristics as "passed pawns", "rook in 7th rank" and so on. My goal is to make the program to find the better strategic path, depending on the chess elements of the actual position. Instead to add a constant value for a passed pawn, for sample, the program must find the right moves to play (strategic path) to promote that pawn. This should be done taking care of all the other elements in the position. I hope to have explained clearly my thought in my poor english ;-))) Ciao!!! Stefano Gemma
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.