Author: Rémi Coulom
Date: 09:23:34 10/23/00
Go up one level in this thread
On October 23, 2000 at 04:43:42, pavel wrote: > >hi Rem > are you still working on TCB, or there wont be anymore TCB in the future? :( > >pavel I have not done any work on TCB for many months and I do not plan to do any in the near future. But I might go back to work on it sometime. The main weakness of TCB is undoubtably its evaluation function. My problem is that I am a rather weak chess player (1500, I would say), and I have trouble to understand positional weaknesses of my program. I am currently doing a PhD thesis in the field of neural networks and I have learnt a lot about TD(lambda) and reinforcement learning in general. Maybe I will try to experiment with such techniques to try to improve TCB's evaluation. It is so bad that it should not be that hard ! I think that reinforcement learning could help also to improve selectivity. When people talk about chess knowledge, they often only think in terms of evaluation function. I believe that selectivity is also an important part of chess knowledge, and can make a big difference. But all of this are just vague ideas, and I am not yet convinced that they might work. Thanks for your interest in my program, anyway. Remi
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.