Newsgroups: comp.ai.alife,rec.games.corewars
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news4.ner.bbnplanet.net!cpk-news-hub1.bbnplanet.com!cam-news-hub1.bbnplanet.com!howland.erols.net!newsfeed.internetmci.com!info.ucla.edu!unixg.ubc.ca!noc.van.hookup.net!vertex.tor.hookup.net!nic.wat.hookup.net!omega.metrics.com!news.maplesoft.on.ca!news
From: Terry Van Belle <vanbelle@maplesoft.com>
Subject: mutant-tolerant redcode
Content-Type: text/plain; charset=us-ascii
Message-ID: <3225BB73.240@maplesoft.com>
Sender: news@maplesoft.on.ca (News File Owner)
Nntp-Posting-Host: agate
Content-Transfer-Encoding: 7bit
Organization: Waterloo Maple Inc.
References: <840378757.19580.0@ogham.demon.co.uk> <4v7hv6$h91@nw101.infi.net> <321EE1F7.42C4@mail.tds.net> <Pine.SOL.3.93.960827084344.14590A-100000@fred>
Mime-Version: 1.0
Date: Thu, 29 Aug 1996 15:46:59 GMT
X-Mailer: Mozilla 2.01Gold (WinNT; U)
Lines: 20

Ross Morgan-Linial wrote:
> Well, the thread over in r.g.corewars is "Artificial Life, Redcode &
> Tierra". They're (we're) trying to figure out a good method of
> representing the redcode instruction set, to be more tolerant of mutants.

I don't subscribe to r.g.c, so I haven't been following the
discussion, but there's no real reason why you need to make Redcode the
language of mutation.  Why not take a cue from nature, where selection
happens on the phenotype, but mutation happens on the genotype?

Design a genotype language which is redundant, fault tolerant, and
completely covers the search space.  Then design a development function
which translates these genotype strings into redcode programs in a
deterministic, many-to-one mapping.  The GA would operate on a
population of genotype strings, and the fitness of these strings would
depend on the performance of their corresponding Redcode programs.
That way you get the benefits of a flexible genetic language and keep
the Redcode specs intact.

Terry
