Newsgroups: comp.ai.genetic
From: stevem@comtch.iea.com (Steve McGrew)
Subject: Re: GAs as a modeling tool?
Organization: New Light Industries, Ltd.
References: <59orp2$ub$1@news.belwue.de>
X-Newsreader: News Xpress 2.0 Beta #2
Date: Thu, 26 Dec 96 15:08:54 GMT
NNTP-Posting-Host: spk0a-11.iea.com
Message-ID: <32c29561.0@news.iea.net>
Lines: 36
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!oitnews.harvard.edu!purdue!haven.umd.edu!news.ums.edu!news.bbnplanet.com!cpk-news-hub1.bbnplanet.com!newsfeed.internetmci.com!news.iea.net!sluggo

In article <59orp2$ub$1@news.belwue.de>, student <tis6evst@rhds01.rz.fht-esslingen.de> wrote:
>Hello.
>
>I have understood how genetic algorithms could be used for
>optimization of highly nonlinear and generally "nonstandard"
>functions.
>I keep wondering however whether GAs can be used as a means
>to build a *model* of some dynamically changing system.
>The questions is formulated rather abstractly. 
>I accept any critics if it is posed inccorect, too.

        Yes, GA's can be used that way.  One way to think of a model is as a 
system of equations, each of which is physically reasonable, which together 
produce the same behavior as the system being modeled.  The objective is to 
design the "simplest" set of equations that produces the desired behavior.

        We have used GA's to develop models, by starting with a mix of 
physically reasonable functions and relationships containing adjustable 
constants and connected by adjustable parameters, designing a fitness function 
that depends on how well the predicted behavior matches actual behavior, and 
adding a penalty for complexity (e.g., the number of nonzero independent 
variables in the model).  This approach gives you a model that not only 
behaves the same as the system being modeled, but also a model that can be 
interpreted and understood.  That is a big advantage over neural networks, 
which typically can accurately match a system's behavior but whose inner 
workings are often completely unrelated to the inner workings of the modeled 
system.

Steve

==================================================================
| Steve McGrew, President    |   stevem@comtch.iea.com           |
| New Light Industries, Ltd. |   Phone: (509) 456-8321           |
| 9713 W. Sunset Hwy	     |   Fax: (509) 456-8351             | 
| Spokane, WA 99204 USA	     |   http://www.iea.com/~nli         |
==================================================================
