Newsgroups: comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!news2.near.net!news.mathworks.com!udel!gatech!swrinde!pipex!uknet!newsfeed.ed.ac.uk!ainews!peter
From: peter@aisb.ed.ac.uk (Peter Ross)
Subject: Re: Q: Name of this GA-operation ?
In-Reply-To: avg@xs4all.nl's message of 26 Feb 1995 19:30:45 GMT
Message-ID: <PETER.95Mar8102954@lyon.aisb.ed.ac.uk>
Sender: news@aisb.ed.ac.uk (Network News Administrator)
Organization: Dept. of AI, University of Edinburgh
References: <3iqkt5$iq7@news.xs4all.nl>
Date: Wed, 8 Mar 1995 10:29:54 GMT
Lines: 35


>  The new operation I'm going to introduce will be a random
>  event ( like mutation ) that occurs after crossover, where
>  certain genes in the chromosome will be exchanged.
>   For example :
>     123456789
>      ^^  ^^
>   will be transformed to  :
>     127856349
>      ^^  ^^
>   I've never read about this operation, but I'm sure this has
>  been done before, and I wonder what name it has in the GA-
>  world.


Exchanging single genes is something I know as `swap mutation',
and use it in work on GAs for timetabling, scheduling and other
constraint satisfaction problems.

I don't know of a name for it when you swap subsequences, but
I'd remark that the same effect is obtainable by four applications
of the little-used `inversion' operator:
  - invert the whole segment from start of first to end of second
    subsequence;
  - invert the middle part to restore it to order;
  - invert both subsequences to restore the order of each.
So you'd need to show that your planned operator is better than
plain inversion. Since inversion seems not to be beneficial in
most people's past experience, it would be nice if this variant
was.

Peter Ross
Dept of AI
University of Edinburgh

