Author: José de Jesús García Ruvalcaba
Date: 14:02:40 04/16/99
Go up one level in this thread
On April 16, 1999 at 17:00:23, José de Jesús García Ruvalcaba wrote: >On April 15, 1999 at 09:27:03, Bruce Moreland wrote: > >>What is "knowledge based" ? >> >>bruce > > Well, I do not know what is "knowledge based". Somebody else wrote a definition >from AI textbooks, perhaps we should adhere to it. > I have tried to figure where lies the chess knowledge of a chess program. I >came up with the following: >1. Evaluation function. This is the first one that comes to mind, the number of >terms, which ones they are, they different weights, and how they change as the >game progresses are some sort of knowledge. >2. Search policy. Where/How much to extend and to prune. A lot of knowledge is >required to unbalance the search tree correctly, without wasting a lot of time >on irrelevant lines of play and without risking an oversight. >3. Time management policy. When to move faster (like when the program faces a >forced recapture, all the other moves being clearly bad), and when to spend more >than average time (like on critical positions). >4. Opening book. Wheter it is automatically produced, or edited by a human >master, the book give the program knowledge to play the opening better >(hopefully) and faster, saving time for the next moves. >5. Endgame tablebases. Strictly speaking they are part of the evaluation >function, but as they provide different information, and most enignes will play >even without them, I list them separately as other form of knowledge. > Of course this is all very simplified, and I can miss a lot of things. >José. I forgot something important: 6. Move ordering. It is vital to have a good move ordering to exploit the search algorithms, which often assume a best case setting. Which moves try first, and for which criteria, is another form of knowledge.
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.