Newsgroups: comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!rochester!cornellcs!newsstand.cit.cornell.edu!portc01.blue.aol.com!newsxfer2.itd.umich.edu!howland.erols.net!vixen.cso.uiuc.edu!newsfeed.internetmci.com!news.ac.net!news.bconnex.net!news.lightlink.com!mv!usenet
From: "Scott Russell" <queuetoo@emrys.mv.com>
Subject: Re: Q: More than one fitness parameter
Message-ID: <01bb8530$54ed8fc0$420116cf@Emrys>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Organization: QC
Date: Thu, 8 Aug 1996 13:47:03 GMT
References: <01bb8343$2091e3a0$LocalHost@Emrys><phwwzc2cc4.fsf@dude.mcs.anl.gov> <4u9kv5$3qt@bignews.shef.ac.uk><01bb8461$1165d0c0$420116cf@Emrys> <ph91bqfnlh.fsf@freedom.mcs.anl.gov>
X-Newsreader: Microsoft Internet News 4.70.1141
X-Nntp-Posting-Host: emrys.mv.com
Content-Transfer-Encoding: 7bit
Lines: 39

> > Different sets of weights resulting in a different solution...
> > Isn't that the goal?  If I simply wanted _a_ best solution, I'd be
> > using linear functions, or some kind of a recursive function.  The
> > reason I'm using GA is so I can find _all_ of the best solutions -
> > or at least a set big enough to satisfy whoever asked for the list.
> > The costs having different units...  Isn't that another simple
> > scaling problem?
> 
> I don't quite understand your question; however, I think *maybe* you
> and I are using the term ``weights'' in two different senses...

I think we _do_ have a communications problem here.  In essence, I'm
following the biological model of genetic programming.  You seem to be
trying to solve linear equasions with ga's, where I'm trying to evolve
organisms that fit certain needs.  

> =BUT=, suppose *you* instead choose to optimize F' = .3*F1 + .7*F2;
> *you* will probably get a *different* set of solutions than I. Are 
> *your* solutions ``better'' than mine, in any sense? In *what* sense,
> and how do you know?  

The conceptual problems we're having stem from our two goals.  A simple
linear algebra problem only _has_ one solution.  It's not fuzzy. You can do
things like reshape or minimize the search space because the problem allows
you to eliminate huge areas of the search off the bat. This is what
Essentially what most 'optimizations' do, and they would suit themselves
well to creating organisms that can F = .5*F1 + .5*F2 quickly.

Instead, if you want an organism that "does not get eaten by the other
organisms,"
and include _lot_ of dynamic parameters, you suddenly have a much larger
portion or a much larger search space available to you.  And instead of
getting _the_ optimal solution, you wind up a number of diverse solutions
that satisfy your goals to one degree or another.

You're looking for 16 = .5*12 + .5*20.  I'm looking for squirrels and
orangutangs.  No wonder we're not looking in the same trees. :)

- Scott
