Newsgroups: comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!uunet!hasler!root
From: root@hasler.uucp
Subject: Re: Random Number Generator Effects
Message-ID: <D8Hpqr.3sI@hasler.uucp>
Organization: Dancing Horse Hill
In-Reply-To: Kent_J_Kostuk@engr.usask.ca's message of 10 May 1995 18:49:01 GMT
Date: Sat, 13 May 1995 00:08:50 GMT
References: <3or1qt$3uq@tribune.usask.ca>
Lines: 23

Kent Kostuk writes:
> I have been using the standard generator in various platforms (Think C, 
> Borland C, and the standard C on  a DEC workstation), and I was 
> questioned on what effect (and how suitable the standard generator was) this 
> this had on my research results.
> Any comments?

Many random number generators are poor.  The rand() in many unix systems, for
instance, is known to suffer from a repeating pattern in the lowest dozen bits.

They are not really random number generators, of course.  They actually only
produce long, repeating sequences with some of the properties of a random
sequence.  What if the number of distinct elements of the sequence is small
compared to the number of solutions in your solution space?  Or your generator
is biased so that a large portion of your solution space is inaccessible?

Maybe we should be using real random number generators.  Anybody got one?

-- 

John Hasler uunet!hasler!root
Dancing Horse Hill
Elmwood, Wisconsin USA
