Lines: 38
Newsgroups: comp.ai.genetic
Message-ID: <5fS7qWEaZNB@bion190.bionic.zer.de>
From: A.FINDEWIRTH@BIONIC.zer.de (Andreas Findewirth)
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!news.sprintlink.net!uunet!zib-berlin.de!uni-paderborn.de!golden-gate.owl.de!royal.owl.de!hsp.zer.de!bionic.zer.de!A.FINDEWIRTH
Subject: Re: Fitness in a competitive population
Date: Tue, 07 Feb 1995 23:00:00 +0000
X-Mailer: CrossPoint v3.02
References: <3h3nu4$5k1@news.bu.edu>
X-Gateway: ZCONNECT UP royal.owl.de [UNIX/Connect v0.71]
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 8bit
X-Z-Post: Eupener Str. 27;D 32051 Herford
X-Z-Telefon: V+49-5221-57764

Karl asked
>B) Does anyone know of a good scheme to estimate fitness in this  
situation
I'm no expert in GAs, but I will try my best.
>1)  The total number of possible strings is 2^N, so only a very small
>    sample can taken, therefor fitness can only be estimated.
Isn't it a normal situation when you use GAs to solve a problem?
>2)  The population changes so the fitness estimate will change over time.
>    This make it a non-stationary optimization problem, which is very hard.
Indeed. But the answer of Nature to this problem was the adaption of  
living systems with techniques as selection, crossing-over and so on.
>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.
Maybe the game players can accumulate something like energy (or money or  
...) by playing a game against each other. The winner takes a fixed  
percentage of the other players energy. So, if A beats 90% of the other  
players, he will win very often, but he gets only a small amount of energy  
in every game. If C can only win against A, the price is much higher, but  
he seldom wins. Isn't that fair? And the number of offspring depends  
(probabilistically) on the amount of energy a player has won.
>4)  Evolutionary stable strategies can arise which collapse genetic
>    diversity.  For example, assume B is a supoptimal strategy, it can only
There are several techniques handling this problem:
- You can use a number of small subpopulation. On every subpop runs a GA  
and from time to time some members of one subpopulation migrates to  
another.
- You can use the technique of dominant and recessive alleles an nature  
does.
In both cases the diversity will be protected.
Excuse me, my English isn't as good as should be
Goodluck
Andreas
------------------------------------------------------------
Andreas Findewirth    Tel.  05221/57764
Eupener Str. 27       EMail A.FINDEWIRTH@BIONIC.zer.de
D 32051 Herford
## CrossPoint v3.02 ##
