Newsgroups: comp.ai.neural-nets
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!newsfeed.internetmci.com!EU.net!peer-news.britain.eu.net!newsfeed.ed.ac.uk!dcs.ed.ac.uk!jura.dcs.ed.ac.uk!iic
From: iic@dcs.ed.ac.uk (Ian Clarke)
Subject: Re: Life with neural net learning?
X-Nntp-Posting-Host: jura.dcs.ed.ac.uk
Message-ID: <DKyqxJ.Gq0.0.staffin.dcs.ed.ac.uk@dcs.ed.ac.uk>
Sender: iic@jura.dcs.ed.ac.uk (Ian Clarke)
Organization: Department of Computer Science, University of Edinburgh
X-Newsreader: xrn 8.00
References:  <mreddy-0901961021260001@mreddy.comp.glam.ac.uk>
Date: Wed, 10 Jan 1996 11:36:52 GMT
Lines: 22

In article <mreddy-0901961021260001@mreddy.comp.glam.ac.uk>, mreddy@glam.ac.uk (Mike Reddy) writes:
> Conway's Life program uses universal rules (0,1 or 4 neighbours is death,
> else life). Langton has quantified Wolfram's four classes by devising
> rules for life/death which conformed to predefined values of lambda
> between 0 and 0.5 (the probability of a cell being alive).
> 
> It occurs to me that CAs are analogous to geographically constrained
> neural networks, where the inputs are the states of neighbours. Has anyone
> done work where the 'rules' for life/death are variable and/or learned?
> e.g. Cell 3c will be alive dependent upon weighted values of its
> neighbours.
> 

I don't know how such a neural network could be trained, as you would
need to specify some desired behaviour, but you might try to evolve
rules for birth/death based upon asthetic or otherwise criterion.

Ian
-- 
|IAN CLARKE        I.Clarke@sms.ed.ac.uk "They couldn't hit an elephant|
|                  iic@dcs.ed.ac.uk      at this dist.."  - Last words |
|I.Clarke@ed.ac.uk ianc@aisb.ed.ac.uk    of General M Howard, 1918     |
