Re: genetic programming

• To: tesla-at-pupman-dot-com
• Subject: Re: genetic programming
• From: "Bill the arcstarter" <arcstarter-at-hotmail-dot-com> (by way of Terry Fritz <twftesla-at-uswest-dot-net>)
• Date: Wed, 14 Jul 1999 11:31:48 -0600
• Approved: twftesla-at-uswest-dot-net
• Delivered-To: fixup-tesla-at-pupman-dot-com-at-fixme

Marco wrote:
>A genetic algorithm is usually faster than a usual hill-climbing one, for 
>problems including also local maximums, but IT IS NOT GUARANTEED TO 
>CONVERGE.

Yes.  This is certainly true, but to be fair *no* search algorithm is 
guaranteed to find the global minimum.

>As any optimum-search algorithm, it is time consuming and that's probably 
>why they use such a computer for it. I ran my genetic algorithms on my home 
>PC (!).

Also true.  However the runtime of a genetic search will increase linearally 
as the number of variables increases.  Most other search algoritm increase 
geometrically with the number of variables.  This is where genetic 
techniques really shine.

-Bill


_______________________________________________________________
Get Free Email and Do More On The Web. Visit http://www.msn-dot-com

• Prev: The magnetic Library (was Re: Tesla stuff for sale)
• Next: Re: Archive Repost - How to bring a dead NST back to life
• Index(es):
  • Main
  • Thread