Author: Graham Laight
Date: 06:29:16 09/23/02
Go up one level in this thread
Ruffian outclassed the opposition quickly and easily by abandonning the current computer chess paradigm, and going for a new one. It is built from off the shelf computing components selected for their speed. When faced with a position to evaluate, it searches it's database of closely matching positions "in memory", and gives each one a score depending on how closely it matches the current position. In the record for the best matching position is an evaluation function, which is not a list of formulae like conventional programs, but a neural network, built with an "off the shelf" neural network package. Each neural network will have been both highly trained and tuned using a commercially available genetic algorithm program. Hence the training of the evaluation is refined in a MASSIVELY more sophisticated way than the evaluation of a traditional programme. Result? A markedly better program, built quickly, simply and easily. Extra ingredient - imagination. -g On September 23, 2002 at 08:00:18, pavel wrote: >Now let me raise the question again. > >Q| Is it possible to write a chess program in such a short time (and not even >visit this, or other, forum) that would play, atleast, at the par of some >commercial engines? > >Answer: Ruffian! > >cheers, >pavs > >ps, Uri, you can safely download it, I checked, there is no virus attached to >it. ;)
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.