Newsgroups: comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!howland.reston.ans.net!pipex!uknet!newsfeed.ed.ac.uk!ainews!dave
From: dave@aisb.ed.ac.uk (Dave Corne)
Subject: Re: Right poulation size
Message-ID: <D6vG33.u7@aisb.ed.ac.uk>
Sender: news@aisb.ed.ac.uk (Network News Administrator)
Organization: Dept AI, Edinburgh University, Scotland
References: <3k0g0l$scm@rover.ucs.ualberta.ca> <3k2qtf$npj$2@mhadg.production.compuserve.com>
Date: Tue, 11 Apr 1995 12:59:27 GMT
Lines: 44


 Salvatore R. Mangano <72053.2032@CompuServe.COM> writes:
>
>In general, you should pick the largest pop size that you can 
>afford for the problem at hand. 

   I'd disagree with this ... the intuition seems valid, of course,
  since with a terrifically large population size you're likely
  to have optima in the initial generation anyway, but this is 
  a very long way down the line!  Generally, we've always found
  a cutoff point -- be it 200, 500, or whatever, after which the
  benefits of increasing the population size become perishingly 
  small when set against the extra time you need to spemd 
  evolving.
>
>It is better to use more run's of fewer generations then 1 very 
>long run.

  Unless your long run involves continual re-initialisation of 
  some sort, perhaps.

>I have had success with smaller populations by adapting the 
>crossover and mutation rates based on the variance in population 
>fitness (or standard deviation). My method is to increase the 
>mutation rate and decrease the crossover rate as the variance 
>decreases and visa versa. Of course, one must not allow the rates 
>to get too large or small.

   I've seen and heard this sort of thing often, and it naturally
  helps things along. You might be better of doing it in terms of
  genotypic rather than phenotypic variance though. 

----------------------------------------------------------------------
   _______       __   Dave W. Corne,
 /   ------.   / ._`_          Department of Artificial Intelligence
|  /         ~--~    \                University of Edinburgh,
| |             __    `.____________________ _^-----^ Forrest Hill,  
| |  I=|=======/--\=========================| o o o | Edinburgh,
  |  I=|=======\__/=========================|_o_o_o_| EH1 2QL, U.K.
 \|                   /                       ~    ~
   \       .---.    .  0131 650 3084  - Dave.Corne@ed.ac.uk
     -----'     ~~''
----------------------------------------------------------------------
    
