Author: Charles Roberson
Date: 08:19:31 02/16/05
Go up one level in this thread
I like the thought of not cloning the personality. But, there may be some issues with the nonpersonality code being copied. For instance, is it ok to copy all the work you did on paralellizing crafty -- I think maybe not. It really saves the person effort, but maybe that was your intent. What if one wanted to test some auto-equation-learning algorithm (ie. neural networks). Then a large series of positions were created from a large number of full games (I mean every position in every game). A Crafty binary or YACE... binary was used to create a series of positions with a numeric eval. An equation-learning algorithm could iterate over the data and learn to mimic the "trainer". In this practice no crafty source could would be needed. Is this acceptable? Is this a clone? These may be two different questions. On February 16, 2005 at 10:46:39, Robert Hyatt wrote: >On February 15, 2005 at 18:38:43, John Merlino wrote: > >> > > >Here is my thinking. > >1. The things that give a program its "personality" are the evaluation and the >search itself. The search defines the program's tactical ability, the >evaluation defines the programs non-tactical chess playing ability. Copying >either/both should not be allowed. > >2. Other parts such as the opening book are probably OK. For example how many >are using a GUI that handles the book outside of the engine? That seems >difficult to stop. Of course a "shared opening book" is a no-no, since that is >another part of a program that defines its playing skill level. > >
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.