Author: Ed Schröder
Date: 03:04:29 04/12/02
Go up one level in this thread
On April 12, 2002 at 05:32:00, sean wrote: >That sounds reasonable to me, I will have to look for the receipt, will let you >know soon. > >I wasn't lying, I was reffering to what Ed posted a while back, > >"It is true, Tiger II has a learning problem. Christophe and Lex are working on >it. The "patch" + an extra increase of playing strength is expected mid January" >Ed > >Patch (a minor usually temporary correction or modification in a computer >program) > >And the learning problem was not fixed in tiger 14> > >"There is a problem with Tiger's learning algorithm, it will be fixed in the >next version" >Ed > >Any way let us hear no more about it and put it behind us. Sean, I am just passing by and saw this by accident. CCC nowadays has so much postings, it's hard to read the ones of my concern. I do remember some discussions on Tiger's learning algorithm, Christophe confirmed the algorithm needed a further improvement and Christophe made it better, it was part of a major free update, Chess Tiger 14. Now realize this, the issue of "learning" is not something fixed such as solving an obvious bug, learning is abstract, as abstract and complex as writing a chess engine and it needs years to develop a good system. So what do we have, in a way it's like testing a chess engine using a large set of test positions (say 100) and when the latest version performs (say 75%) and the older version (say 60%) the latest version is better. I am sure you get the point, the 25 positions the newer version still does not solve are not a convincing argument saying, "the bug hasn't been solved". Learning is a complex issue resulting in complex and abstract software, it certainly is not a kind of typo that can be fixed in a couple of minutes. I hope my words have spread some light on the issue. My best, Ed
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.