Author: Uri Blass
Date: 08:44:41 12/29/01
Go up one level in this thread
On December 29, 2001 at 07:48:38, Tom Kerrigan wrote: >On December 29, 2001 at 07:44:19, Sune Fischer wrote: > >>On December 29, 2001 at 07:24:06, Tom Kerrigan wrote: >> >>>There is nothing in my idea that is specific to tables. >>>-Tom >> >> >>Must have been all the talk about the knight on h8:) >> >>Still you dimensional search space is very large, and with each concensus eval >>that expensive, you can't really afford a sloppy approach. >>Genetic algorithms might do the job if you can define a good consensus/scoring >>function, which is no trivial thing to do in the first place. >> >>-S. > >I think we're talking about different algorithms. My idea is for making a >program play a known-good move, not to find ideal values for all eval weights. >The hope is that if you force the program to make enough good moves, the eval >weights will end up being pretty good. > >-Tom It is similiar to the idea that I suggested when the idea was to test a change by a test suite if the program avoid more bad moves http://www.icdchess.com/forums/1/message.shtml?203666 The differences are: 1)I talked about tactical mistakes that can be verified by a long search(I agree to add positional mistakes if you can be sure that the move is a positional mistake) 2)I did not suggest to look at one position in order to decide if to change the evaluation because it is possible that you create bigger problems by correcting one problem. I do not believe that it is possible to find all good moves(or avoid all bad moves) by changing the weights of the evaluation function and this is the reason that I do not like the idea of deciding about a change in the evaluation based on only one position because I am afraid that you may destroy a lot of correct evaluations by forcing the program to evaluate correctly one position. Uri
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.