Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Puzzled about testsuites

Author: Michael Yee

Date: 13:20:21 03/09/04

Go up one level in this thread


On March 09, 2004 at 16:06:47, Gian-Carlo Pascutto wrote:

>On March 09, 2004 at 15:37:30, Michael Yee wrote:
>
>>Additionally, I think there could be even less danger in the case of chess since
>>the training data has no error (in theory).
>
>What?!

What I mean by "no error" is that technically there's no "measurement" error
like in other applications (where you might apply regression). For example, if
you wanted to predict crop yields based on amount of rainfall, you'd expect the
measurements in the training data to have some measurement error (like guassian
noise). But the chess examples are discrete. And if you chose to assume that a
consensus of GMs was the "truth", where would the random error be?

Michael



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.