Author: Vincent Diepeveen
Date: 13:44:18 08/05/02
Go up one level in this thread
On August 04, 2002 at 15:24:50, Sune Fischer wrote: >On August 04, 2002 at 14:46:18, Vincent Diepeveen wrote: > >>On August 04, 2002 at 14:06:55, Sune Fischer wrote: >> >>will, believe, think, consider. >> >>Please proof it. chess is very simple compared to other >>applications where automatic tuning is supposed to work >>in the future. >> >>So far i have not seen a single decent program that >>can do better with automatic tuning than without. > >Well, maybe you have, do you know how the pros are tuned? Yes they all do it by hand. Most even tell me they also tried autotuning but it failed miserably. >>there is a shitload of freeware programs and volunteers to rewrite >>them to enable automatic tuning. Please >>pick a strong program and tune it. I would advice crafty. >>A small parameter set. Even big advantage for the tuners, >>but already a good program to start with. >It is a non-trivial exercise to do, and I don't know every character of >Crafty's code. Besides I would rather spend time on implementing this in my own >program and get an edge :) The hard truth is that there only exists a few 'academic' approaches which all are complete overoptimistic presented results. In tuning chess programs there has not been a single success in absolute terms. No Knightcap didn't play better after tuning. It played worse. Yet the fact that he managed to make A FORM OF TUNING, it means he already did something great, because majority is talking about potential of it without realizing what they say! >I believe I can make it work, maybe even improve on it. >It isn't real important to me whether you believe it works or not, I think you >should follow your ideas and I will follow mine, actually I prefer if you forget >all about TDLeaf as soon as possible :) I didn't forget it of course. In fact i've had whole courses that have to do with neural networks. Where i had to hear the same crap from someone who doesn't know a thing on real world applications. the hard facts in game playing world is simple: we need accurate values. You don't design a pattern in order to let some 'auto' tuner fuck it completely by giving a -y instead of a +x value to it. That's a waste of time of designing such a pattern then! >>Finding the best values as a human isn't trivial. It sure isn't >>for programs. But humans use domain knowledge your tuner doesn't. > >KnightCap was too interesting a project not to follow up on, I'm very surprized >it hasn't been done already. >To see people write that it doesn't work when a) KnightCap proved it _did_ work, >and 2) they have not even attempted it themselfs, is very funny to me. Understand me wrong. i am not saying he wasted his time. Instead i feel he did a good job. Nevertheless, let it be clear that he achieved nothing concrete. He managed to make something and it didn't work. that's usually a more valuable source of information than someone who is just 'guessing' something doesn't work. Of course like all scientists he wasn't allowed to mention it didn't work, so he started with some kind of random set and then presented the learned set. That means it at least *improved* itself upon random values. Nevertheless for me a learner only works if slowly better and better local maxima in the parameter set are getting found. This is a major problem for all learners so far. Not a single learner that is not somehow brute forcing its learning is capable of slowly improving its local maximum. Amazingly not a single scientist who made tuning so far has done effort to even *try* to show that each so many testsets the local maximum improved. That says itself enough about the quality of the research. It says even more about the usefullness for real world applications where accuracy is required! >-S.
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.