Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!think.com!samsung!munnari.oz.au!bruce!monu0.cc.monash.edu.au!monu4!ins845b
From: ins845b@monu4.cc.monash.edu.au (mr  k.l. lentin)
Newsgroups: comp.ai.neural-nets
Subject: games and pattern recognition
Message-ID: <1991May22.035830.27821@monu0.cc.monash.edu.au>
Date: 22 May 91 03:58:30 GMT
Sender: news@monu0.cc.monash.edu.au (0000-news(0000))
Organization: Monash University, Clayton, Vic 3168, Australia
Lines: 16

As part of a computer science project, we are trying to build a net that
will be able to play a game. The ultimate aim could be as high as chess but
for the moment we are content to consider naughts and crosses (tic-tac-toe)
I would appreciate some advice on where to start or maybe a paper on game
playing.

Also, as a way of learning to use nets, we have been trying to reognise some
patterns. To keep the games flavour, we have been dealing with detecting and X
or an O on a retina. We wouild like the X or O to be reccognised no matter 
where it is  or how big it is. We have tried a simple back propagation net and
a Hopfield-Net but both recognise only what we train them with, they can't
detect the different positions.
Again any help here would be much appreciated.

|/
|\evin