Author: Sune Fischer
Date: 00:38:03 10/26/02
Go up one level in this thread
On October 25, 2002 at 19:38:24, Ingo Lindam wrote: >On October 25, 2002 at 14:49:24, Sune Fischer wrote: > >>I think sometimes things take a while to sink in, even with researchers. >>All these ideas may look good on paper, but there are some hefty pratical >>obstacles. Consider the resolution on such an evaluator for instance, let's say >>this NN is all the positional evaluation you have (and then some material bean >>counting). Now you search 500 kNps and do maybe 100 kNps full evaluations (just >>to throwing out some numbers). If you search for 10 secs that's 10^6 >>evaluations, and you probably have not gone more than ~10-12 plies which means >>most of the positions (except those few heavily extended) are going to be very >>similar and extremely hard for a NN to distinguish, you really need an awsome >NN and with that kind of power better tricks are probably at your disposal. > >Hello Sune, > >I am not sure what I (or who) said misleading you to guess that I was thinking >of a neural network. I was was thinking of a statistical approach using a more >or less conventional search algorithm. Ofcourse I hope the statistical pattern >allow a better pruning. Sorry about that :) Must be the word "patterns" that trigger a lot of associated knowledge. >For the generation of the pattern I thought of a very different data structure >to represent the games. Spending more memory then e.g. cb-format but allowing a >very efficient access by the lowest level pattern of a position. But here is the problem: "the lowest level pattern of a position", what does that mean exactly? You want to pass the whole position into some grinder and get out an evaluated score, right? Usually this is done by identifying small patterns, eg. pawn structure around the king. I don't see how you are going to make a very efficient access without looking specificly for individual patterns? -S. >Best regards, >Ingo
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.