Author: Steven Edwards
Date: 00:51:28 06/24/05
Go up one level in this thread
On June 24, 2005 at 00:47:53, Roger D Davis wrote: >Sounds like you're getting a lot closer. Will be fascinating to watch Symbolic >play it's first game with the planners in place. Can't wait. :) Thank you for your comments. I encourage patience as the end is still a long way away. A much closer goal is to have Symbolic handle long nontrivial mating sequences via the simple planning that uses the GA derived microfeature recognizer. The next goal after that will likely be using more advanced pattern recognition that allows better and longer plans, although the domain will still be restricted to mating sequences. Around that time it will become possible to validly compare Symbolic's mate solving performance on test suites with traditional A/B searchers. Once the above is done, I might take some time off to do a paper on the project. Two reasons: first, to encourage others to consider a cognitive approach; second, to demonstrate that Symbolic is a real, functioning program and not a paper-only program with unsubstantiated claims. (The cognoscenti will know the reference.) Afterwards, the task is to implement general tactical planning within the existing framework. With this, Symbolic's processing can be compared with the output of Berliner's CAPS and Wilkins' Paradise on selected problems. Eventually, the program will incorporate extensive knowledge about positional goals and strategic themes. At that point it should be capable of playing a decent (Elo 1800) game using only the Lisp cognitive search without any assistance from the toolkit's traditional A/B routine.
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.