Newsgroups: comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!news.alpha.net!uwm.edu!vixen.cso.uiuc.edu!howland.reston.ans.net!Germany.EU.net!EU.net!sun4nl!news.nic.surfnet.nl!tudelft.nl!news.twi.tudelft.nl!boi
From: boi@dutiws.twi.tudelft.nl (Boi Sletterink)
Subject: Fitness value range and semantics
Organization: Delft University of Technology
Date: Fri, 24 Feb 1995 10:24:21 GMT
X-Newsreader: TIN [version 1.2 PL2]
Message-ID: <D4I28n.4xo@dutiws.twi.tudelft.nl>
Lines: 25

Hi,

I'm currently working on a PGA for partitioning non-regular grids into equally
sized partitions. There's just one tiny detail that keeps bugging me (it
doesn't keep me from my work, but I still wonder...):

A fitness function is always a positive number, and higher numbers indicate
a better solution. 

So far, so good. Now I've got a fitness function that depends on the communi-
cation cost between the partitions (cq. processors); this communication cost
is of course, always a positive number, but large numbers indicate worse
solutions. So, I subtract it from the maximum communication cost to make it
conform the fitness specs, because something like that is mentioned in
literature; I've never seen an alternate approach.

My question is: are there any objections about turning the PGA problem into
a minimizing problem in stead of a maximizing problem (besides the somewhat
funny meaning of the fitness value) or allowing negative fitness values
(so that I don't have to find this maximum communication cost, but higher
numbers still indicate better solutions)?

Thanks in advance for any response,
  Boi Sletterink

