NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/sci/math/numanal/2261

Text File | 1992-07-21 | 3.7 KB | 77 lines

Newsgroups: sci.math.num-analysis
Path: sparky!uunet!sun-barr!cs.utexas.edu!usc!sol.ctr.columbia.edu!The-Star.honeywell.com!umn.edu!noc.msc.net!uc.msc.edu!apctrc!zkdc05
From: [e-mail] (Kelly D. Crawford)
Subject: Re: Choice of optimization methods
Message-ID: <[e-mail]>
Originator: zkdc05@church
Sender: [e-mail]
Organization: Amoco Production Company, Tulsa Research
References:  <[e-mail]>
Date: Tue, 21 Jul 1992 22:15:06 GMT
Lines: 64


In article <[e-mail]>, [e-mail] (JULIO RODRIGUEZ BANGA) writes:
>
>   There is a huge amount of literature reporting the solution of optimization
> problems using genetic algorithms and direct search optimization methods (like
> the Complex from Box) that are more or less 'stochastic'.
>
>  I'm myself a user of this type of methods because they are usually easy to
> work with and, more important, they usually work well, if you don't mind CPU
>  time very much.
>
>  However, these methods are usually regarded as 'politically incorrect', and
> considered as 'inferior' when compared with, for example, SQP (sequential
> quadratic programming) and gradient methods. For those interested, see the
> comments of Dr. Sargent in the book 'Foundations of computer-aided process
> design', edited by Sirola, Grossman and Stephanopoulos, CAChE-Elsevier, 1990.
>
>  Though my work deals with optimization, I'm not an expert in the field, so I'd
> like to receive opinions about the pros and cons of GA and direct search
> methods when compared with 'politically correct' algorithms. Particularly,
> if those methods are 'incorrect', why are they so widely used?.
>

Let me offer you a summary given by Lawrence Davis at the Fourth International
Conference on Genetic Algorithms.  It is one of the best ways I've found of
determining when to use a GA:

     When to Use Genetic Algorithms

     - Problem requires good, but not optimal solution
     - Acceptable performance measure is available
     - Feasible to test many potential solutions
     - Acceptable representation (of a solution) is available
     - Size and complexity of search space preclude traditional approaches:
          - Analytic solution
          - Exhaustive search
          - Hill-climbing
          - Random search

I don't know about 'political correctness', but as you have already pointed out,
the technique works and is widely used.  My input is that if your problem satisfies
the above criteria, you should consider a GA.

If someone yells at you for using a GA when it happens to be a good way to
solve your particular problem (the GA is out looking for mountains while their
gradient technique converges to a local optimum on a small hill), then you should
ask them to provide a better solver to back up their words.  Unless, of course, it
happens to be your boss :-)

Seriously, though, GAs have ample theory to suggest why they work, and although
there is a certain degree of randomness (in fact, a LOT of randomness), there is a
high probability that they will converge toward a good solution.

Can you elaborate about what you mean by saying that "these methods are usually
regarded as 'politically incorrect'"?  I know that many people don't understand them
and tend to shy away from them because of their unfamiliarity (everyone does this
with something), but I don't think I've ever heard them called 'politically incorrect'.

Kelly
--
Kelly D. Crawford             | Amoco Production Research | These are my own
[e-mail]       | 4502 East 41st Street     | opinions, not
phone: [phone]         | Tulsa, OK  74102          | Amoco's...
          >>> *** This line intentionally left blank *** <<<