Newsgroups: comp.ai.alife,comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!rochester!cornellcs!newsstand.cit.cornell.edu!news.kei.com!newsfeed.internetmci.com!EU.net!peer-news.britain.eu.net!newsfeed.ed.ac.uk!ainews!andrewt
From: andrewt@aisb.ed.ac.uk (Andrew Tuson)
Subject: Re: mutation/sexual reproduction (was Re: Has anyone written a genui...)
Message-ID: <DJ2AxA.D5y@aisb.ed.ac.uk>
Sender: news@aisb.ed.ac.uk (Network News Administrator)
Organization: Dept AI, Edinburgh University, Scotland
References: <49e537$abg@news.express.co.nz> <49g7ev$4sq@news.onramp.net> <DJ28u3.n2y.0.staffin.dcs.ed.ac.uk@dcs.ed.ac.uk>
Date: Mon, 4 Dec 1995 12:34:22 GMT
Lines: 72
Xref: glinda.oz.cs.cmu.edu comp.ai.alife:4731 comp.ai.genetic:7424

In article <DJ28u3.n2y.0.staffin.dcs.ed.ac.uk@dcs.ed.ac.uk> bjm@dcs.ed.ac.uk (Bruce McAdam) writes:
>[Kinda messily trying to follow up to Peter Harrison's original message and
> Steve Dekorte's followup]
>
>> It seems to me that combining the gemones of individuals is just an 
>> *additional* form of mutation with obvious benefits. 

So, what are they?

>> Understanding this and seeing that 2 genome 'sex' is popular in nature 
>> for physical and not informational reasons leads us to ask why combining 
>> >2 genomes is not a more popular form of mutation for GA/AL algorithms(?)

Erm, I could well be misunderstanding this, but wasn't crossover in GAs
inspired by sexual reproduction?

>Evolution works neither by combining organisms' geneotypes or by mutating 
>genotypes but by selecivly preserving the different variations in genotypes
>that arise by ANY means.
>
>Sexual reproduction has not arisen because it is successful at creating
>complex organisms but because DNA which makes organisms reproduce survives
>better. 

IMHO, I don't have a great deal of truck with biological metaphors
for GAFOs (GA as Function Optimiser), though Alife stuff is a different
story...

The reason that sex probably evolved (its still an active argument!) in
nature, is different from why crossover is thought to be a good idea in GAs..

To summarise, when you look at the problem closely enough, it is a wonder
why sex evolved at all!

I would recommend reading:

The Red Queen: Sex and the Evolution of Human Nature, by Matt Ridley
The Evolution of Sex, by John Maynard-Smith

For an overview of the 'why-did-sex-evolve' debate.

>I would say it is likely that complex organisms (like ourselves) could not 
>have arisen without sexual reproduction so it is probably a good idea to use
>it primarily but mutation is also useful to keep the gene-pool topped up 
>with different combinations of genes which have either not arisen before of
>were not found useful before.  (I think research has shown that a GA can work
>without any mutation).

Yes, GAs can 'work' in a fashion without mutation, but solution quality
suffers as a result (you are restricted to the diversity present in the
population, and the GA cannot be guaranteed to find the optimum).

GAs can work without crossover, and many simple organisms evolve quite happly
without sex!

(From an optimisation viewpoint, is a population really necessary anyway?
Simulated Annealing seems to do fine without one. BTW, I don't want to start
another GA vs SA debate....)








-- 
Andrew Tuson (andrewt@aisb.ed.ac.uk)

Department of Artificial Intelligence, University of Edinburgh, Scotland, U.K.
An expert is a person who avoids the small errors while sweeping on to the
grand fallacy..........:-)
