Newsgroups: comp.ai.alife
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!udel!gatech!howland.reston.ans.net!pipex!uknet!bcc.ac.uk!link-1.ts.bcc.ac.uk!zcaccha
From: charris@cs.ucl.ac.uk (Christopher Harris)
Subject: Re: Evolution Maker
Message-ID: <1994Nov14.140813.21400@ucl.ac.uk>
Sender: zcaccha@link-1.ts.bcc.ac.uk (Christopher Harris)
Date: Mon, 14 Nov 1994 14:08:13 GMT
References: <Cy8M95.8t9@udcf.glasgow.ac.uk> <1994Nov2.104653.26949@leeds.ac.uk> <39oc58$gla@news.duke.edu>
Organization: University College London
Lines: 24

|> : experiences in this area.  Previously I have used asexual reproduction, relying
|> : purley on mutation to produce new organisms but perhaps using sexual
|> : reproduction might speed things up?
|> 
|> Greetings.
|> 	I would like to comment on this.  From what I have observed, 
|> evolution (in this case the emergence of new traits and the refinement of 
|> existing ones) occurs very slowly when your only source of genetic 
|> variability is the introduction of random 'point' mutations (i.e. bit 
|> flips or base changes).  To really hustle things along, and to help skirt 
|> around the problem of local optimization, you need to introduce 
|> large-scale genomic rearrangements and exchanges.
|> 	I would code things like inversions, deletions, and recombination 
|> into your model.  I'm sure this has been done by others--what do people 
|> feel about this?
|> 
|> 			Dan Tomso
|> 

Absolutely! Mutation is a very small factor in the variability of a population, and
is often over-estimated in its power. Crossover and other operators are far more useful,
given a long enough genome and sufficient population.

Chris
