Newsgroups: comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!swrinde!newsfeed.internetmci.com!in2.uu.net!pokey!glendora!m1mbg00
From: m1mbg00@newfed.frb.gov (Michael B Gordy)
Subject: Re: GA 's in MATLAB
Message-ID: <Dnuq78.HLv@glendora.uucp>
Sender: usenet@glendora.uucp (Usenet)
Reply-To: m1mbg00@frb.gov
Organization: Federal Reserve Board
References:  <epohl.27.313AED52@AFIT.AF.MIL>
Date: Wed, 6 Mar 1996 15:11:32 GMT
Lines: 36

In article <epohl.27.313AED52@AFIT.AF.MIL>, epohl@AFIT.AF.MIL (Edward Pohl) writes:
> Does any one know of any GA code written in MATLAB?  I am interested in using 
> GA's and SA for some problems that I already have coded in MATLAB.  I know I 
> can start from scratch, but why reinvent the wheel?

Good point.  A few months ago I posted the m-function below.  Immediately
after my signature is the "README," and then the m-function follows.
Suggestions welcome, but the code is *not* formally supported.

-- Michael Gordy
   Federal Reserve Board

------------------------- ga.README ---------------------------------------
GA.M:  A Matlab routine for function maximization using a Genetic Algorithm.

 The Genetic Algorithm (GA) is a essentially a search-based method of
 function maximization, in which the search over the parameter space
 is guided by evolutionary mechanisms ("survival of the fittest").

 Relative to other methods, the GA has stengths and weaknesses:

 Strengths:
 1. The GA is especially useful for functions with kinks and 
    discontinuities, because it does not make use of the gradient.
    It also performs well when there are lots of local maxima to
    avoid. 
 2. The GA is handy when parameters must be constrained to a compact set.
    It also does not require the user to specify a "starting point."
 3. The GA is quite robust.  Unlike gradient-based methods, it has no 
    tendency to "blow up" if it starts too far away from the optimum.
 4. Relative to other search methods, the GA is fairly efficient in
    high-dimensional problems.

 Weaknesses:
 1. When a gradient function is available and well-behaved, the GA
    is less efficient than me
