Newsgroups: comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!scramble.lm.com!news.math.psu.edu!news.cse.psu.edu!rutgers!sgigate.sgi.com!swrinde!howland.reston.ans.net!newsfeed.internetmci.com!news.kei.com!nntp.coast.net!harbinger.cc.monash.edu.au!newshost.anu.edu.au!csis!news
From: John McLaughlin <John.McLaughlin@cbr.dit.csiro.au>
Subject: selection technique
Content-Type: text/plain; charset=us-ascii
Message-ID: <31C4F990.41C6@cbr.dit.csiro.au>
Sender: news@cbr.dit.csiro.au
Nntp-Posting-Host: noosa
Content-Transfer-Encoding: 7bit
Organization: CSIRO
Mime-Version: 1.0
Date: Mon, 17 Jun 1996 06:22:08 GMT
X-Mailer: Mozilla 3.0b4 (X11; I; IRIX 5.3 IP22)
Lines: 19

Hi,

This is probably extremely old news, but I thought I'd post
a method I came up with for selecting parents from a population.
I keep my population in a list sorted in order of fitness, where
p[0] is the most fit individual. Say there are n individuals,
to non-deterministicly select a fit individual, I choose

p[ random(n) * random(n) / (n-1) ]

Where random(n) is a random number in the range [0,n-1].
Note that there is a bias toward low indices, but any
individual has the potential of being chosen.
So, for my current GA project I select two parents with a
low (fit) bias and one child to replace with a high (unfit)
bias, then perform crossover and mutation. I like it
because it's fast.

-John McLaughlin
