Author: Chris Carson
Date: 16:36:12 06/23/01
Go up one level in this thread
On June 23, 2001 at 17:17:10, Mogens Larsen wrote: >On June 23, 2001 at 16:45:03, Chris Carson wrote: > >>I think I understand what you mean, but I am not sure. Do you mean by "nuances" >>the ability of the engine instead of the ability to the entire program package? >>I would agree that an engine with no database support is a lot weaker. If you >>include learning databases (created from only engine play) then it would take a >>lot of games to move back into the 2500+ ratings range (but it would be possible >>if enough games are played). > >That would be one "nuance" to consider. There's essentially nothing wrong with >trying to prove that a complete package is of GM strength. But is this package >playing computer chess or advanced chess due to the major contribution by GMs >and Nalimov? The answer isn't straigtforward IMO even though some may disagree, >because it's all created by humans nonetheless. > >Even so, I believe there's a cognitive difference between creating an engine >that retrieves external knowledge by itself and adding external sources. Mainly, >because it affects the development of the engine, ie. less need for investing >time in opening and endgame knowledge. If there was effective middlegame >databases they would be used indiscriminately. I don't particularly like that >idea. > >>My program "Dallas" has a very simple eval with very fast searching (mostly >>written in assembly) and builds a learning database from playing games and it's >>analysis of games, it updates the opening book, middle and endgame knowledge >>databases during this period. This means that changing the engine may make the >>existing database mostly worthless, but Dallas has captured a lot of knowledge >>this way, knowledg that I did not have to program into it, thus reducing engine >>changes to bug fixes for the most part. Anyway, I digress. :) > >That illustrates some of the possibilities. Another is TD (temporal differences >I think) used in EXchess, where piece values and other parameters are adjusted >from game to game. Combine with learning databases and you have something >interesting IMO. > >Regards, >Mogens You bring up some good points that I will think about. :) Best Regards, Chris Carson
This page took 0.01 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.