Newsgroups: comp.ai.genetic
Path: cantaloupe.srv.cs.cmu.edu!das-news2.harvard.edu!oitnews.harvard.edu!purdue!lerc.nasa.gov!magnus.acs.ohio-state.edu!math.ohio-state.edu!news.cyberstore.ca!skypoint.com!umn.edu!news
From: "James Albert Larson" <larso171@maroon.tc.umn.edu>
Subject: Sugal V2.0 User Group (one person user group)
To: thasan@mayfair.demon.co.uk
Message-ID: <19955.larso171@maroon.tc.umn.edu>
X-Minuet-Version: Minuet1.0_Beta_16
Sender: news@news.cis.umn.edu (Usenet News Administration)
Nntp-Posting-Host: dialup-2-119.gw.umn.edu
X-Popmail-Charset: English
Organization: University of Minnesota, Twin Cities
Date: Wed, 30 Aug 1995 06:59:48 GMT
Lines: 46

Thasan,

I wrote Dr. Anthony Hunter today with praise on Sugal V2.0 and suggestions, 
namely that he include a Windows .exe executable at his site so that people 
can run it without compiling it first.  This would probably be an .exe 
version of the Dejong function set F1 through F5 that is included in the 
original package as source code.  One can get a lot out of playing with the 
Dejong functions, e.g. real vs. integer vs. bit representations, varying 
mutation parameters, varying the mutation type (e.g. from bitwise to 
gaussian mutation of integers or real numbers), varying the crossover 
method (one point, 2 point, uniform).

Note that in order to model one's own problem, one must be able to compile, 
because even to add one's own evaluation function requires compilation.

I also found what looks like an error in the Dejong F2 function -- there's 
a "+" sign where Goldberg has a "-" sign.  

I also requested support for multiple parallel genetic operators (e.g. 
where the user specifies a 20% probability of choosing crossover operator 
1, a 25% probability of chossing crossover operator 2, ... a 15% 
probability of choosing mutation operator 1, a 20% probability of choosing 
mutation operator 2, ..., and so on.  Currently, the Sugal algorithm allows 
the user pre-run to choose one among several crossover operators and one 
among several mutation operators.  But once that one crossover operator and 
one mutation operator is selected pre-run, the Sugal algorithm uses only 
those two operators during the run.  I believe this is a serious weakness 
that prevents it from solving serious problems.  (On the other hand, Sugal 
is very highly extensible, so I believe one can add this capability on 
one's own without affecting or changing a lot of code, something I will 
shortly be looking at).

I also enclosed some writeups to Dr. Hunter: (1) On how Dejong function 
F1 variables is represented in the bit representation, including the 
operation of the SuGetInteger() and SuGetUnsignedInteger() functions.  (2) 
On how SuSetParameter() works, including diagrams of the data 
structures involved and how they are linked.  (3) The SuPool data stucture 
and how the pool data structure elements are linked.  These documents are 
all in Ascii text (a great feat of Ascii art -- "drawn" in Excel using a
column width 1 spreadsheet, and then saved as ascii text).

I still haven't looked into my own ftp site.  I'll be  asking about it at 
the same time I post this.

Jim Larson

