Newsgroups: comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!hookup!swrinde!howland.reston.ans.net!news.sprintlink.net!redstone.interpath.net!hilbert.dnai.com!nbn!news
From: Ron Macken <rmacken@calon.com>
Subject: Re: Fitness in a competitive population
Message-ID: <D3o1n9.734@nbn.com>
Sender: news@nbn.com
Organization: North Bay Network's news posting service - not responsible for content
References:  <3h3nu4$5k1@news.bu.edu>
Date: Wed, 8 Feb 1995 05:23:32 GMT
Lines: 31

meissner@space.bu.edu (Karl Meissner) wrote:
>
> 
> 3)  The arise of voting paradoxes confounds the attempt to express fitness
>     as an analog number.  For example given strings A, B and C. 
>     A beats B, B beat C, C beats A.  What is their fitness? The same?
>     What if A can beat 90% of the other strings but always loses to C and
>     C can only win against A.

> .....

> B) Does anyone know of a good scheme to estimate fitness in this situation

It's difficult to comment without understanding the game involved but
here's a shot:  At the beginning of each generation, randomly assign
strategies to different groups.  Each group plays a single elimination
tournament.  Each group tournament winner survives until the next 
generation.

This method of selection is fairly brutal and many good strategies will
be eliminated by having the bad luck of facing a tough opponent.  But the
best strategies will eventually rise up and it may help your problem of
convergence.

As far as your concern about "C can only win against A", the evolutionary
process almost always eliminates brittle, specialized solutions.  An
"improved C" will arise that doesn't have the limitations of "C".

It sounds like an exciting project.  Good luck!

Ron Macken
