
Genetic Algorithms Digest   Thursday, June 13 1991   Volume 5 : Issue 15

 - Send submissions to GA-List@AIC.NRL.NAVY.MIL
 - Send administrative requests to GA-List-Request@AIC.NRL.NAVY.MIL

Today's Topics:

	- ICGA-91 Workshop:  Artificial Life and GA Approaches
	- Re: TSP
	- Re: Gray codes
	- Re:  Genesis ftp site?  (2 messages)

**********************************************************************

CALENDAR OF GA-RELATED ACTIVITIES: (with GA-List issue reference)

4th Intl. Conference on Genetic Algorithms (v5n9, v5n14)     Jul 13-16, 1991
AAAI 91, National Conference on AI, Anaheim, CA              Jul 14-19, 1991
EUROPEAN SUMMER SCHOOL on MACHINE LEARNING (v5n7)            Jul 22-31, 1991
Genetic Algorithm Course, Stanford (v5n11)                   Jul 22-26, 1991
IJCAI 91, International Joint Conference on AI, Sydney, AU   Aug 25-30, 1991
First European Conference on Artificial Life (v5n8)          Dec 11-13, 1991
ECAI 92, 10th European Conference on AI (v5n13)              Aug  3-7,  1992

(Send announcements of other activities to GA-List@aic.nrl.navy.mil)

**********************************************************************
----------------------------------------------------------------------

From: todd@galadriel.stanford.edu
Date: Tue, 11 Jun 91 13:06:21 PDT
Subject: ICGA-91 Workshop:  Artificial Life and GA Approaches

    Genetic algorithms and other evolutionary simulation methods have an
    important place in the emerging field of artificial life research.  This
    workshop will explore the applicability and place of evolutionary methods
    in artificial life projects, the features these methods should embody, and
    the successes (and failures) of current ALife research programs that use
    an evolutionary approach.  We look forward to a lively discussion and
    exchange of ideas in this forum.  If you would like to present a short
    rundown of your research and thoughts, or just join in the discussion, or
    if you have ideas about topics you would like to see addressed, please
    send me a message as soon as possible, and I will add you to our mailing
    list.  A tentative schedule will be sent out when it becomes clear.

          cheers,
          Peter Todd
          Stanford University Psychology Dept.
          todd@psych.stanford.edu

------------------------------

Date: Thu, 30 May 91 08:40:00 MDT
From: whitley@CS.ColoState.EDU (darrell  whitley)
Subject: Re: TSP

    Mark Lidd recently posted a message that the results for the TSP which
    we published in the 1989 GA Conference are not the best known.  This is
    correct.  We noted in the paper that the results might be due to rounding
    error--which is in part correct:  our evaluation funded rounded to
    integers before summing the tour length, while other evaluation functions
    round to integers after summing the tour length.  

    We noted these corrections in a 1990 journal article published in
    JETAI and in the Chapter that appears in Dave Davis' "Handbook of
    Genetic Algorithms."  In both publications we noted that when
    larger populations are used that edge recombination matches the
    best known on 30 and 105 city problems with considerable reliability.
    (Our most current results:  finds best known 100% of the time on
    30 cities, finds best known 50% of the time or better on 105 city
    problems).

    In studying recombination for TSP problems more closely, it is clear
    that one of the problems is the alphabet size required for
    permutation problems.    One must use relatively larger populations
    to ensure good representations of all possible edges.    This may
    explain why certain researchers continue to report good results
    using high levels of mutation--mutation introduces edges that may
    be missing in modest size populations (e.g. we had to use a population
    size of 5000 for a 105 city problems using just recombination--no
    mutation.)    It is also likely that Goldberg's work with
    "virtual alphabets" is very relevant to permutations problems using
    symbolic encodings.   

    Ulder presented results using a hybrid GA/Traditional methods on problems
    up to 666 cities.  I believe the challenge for TSP results now depends
    on two issues:  solving larger problems (e.g. 500 cities or more) and
    reliability (finding the best known solutions some reasonable
    percentage of the time). Ulder's paper appeared in "Parallel Problem
    Solving from Nature."  

    The differences in the evaluations functions were first pointed
    out by Fred Glover and Dave Schaffer a few months after the conference.

    Darrell Whitley

------------------------------

From: pl160988@mtecv2.mty.itesm.mx (Ing. Ivan Ordonez)
Date: Wed, 29 May 91 18:08:38 CST
Subject: Gray codes

	    Everybody seems to be talking about Gray codes now. As a small
    contribution, I am sending two algorithms for converting binary to Gray
    and Gray to binary, writen in Pascal-like pseudocode. Let Chrom be a
    type defining an array of binary numbers, the first element being the
    less significant one, and let lchrom be the number of bits in the array.

    Procedure GraytoBin (Var cGray, cBin : Chrom)

    for j := 1 to lchrom - 1 do
	    cGray [j] := cBin [j] XOR cbin [j + 1]
    cGray [lchrom] := cBin [lchrom]

    Procedure BintoGray (Var cBin, cGray : Chrom)

    value := cGray [lchrom]
    cBin [lchrom] := value
    for j := lchrom - 1 downto 1 do
	    if cGray [j] = 1 then value := NOT value
	    cbin [j] = value


	    These algorithms are very efficient, because they run in linear
    time (an algorithm involving matrix multiplication would take at least
    cuadratic time).

						    - Ivan
------------------------------

From: Alan C. Schultz - GA-List Moderator (schultz@aic.nrl.navy.mil)
Date: Sun, 9 Jun 91 21:03:19 EST
Subject: Re:  Genesis ftp site?

I have received several submissions to ga-list recently concerning
the availability of GENESIS, and these discussions have led to several
readers asking questions like the following, submitted by
John C. Schultz (no, we are not related!)...

>>    Could someone explain how [Grefenstette's] GENESIS is different than
>>    the one whose partial UsersGuide I have included below?
>>
>>                 GENESIS 1.2ucsd was derived from  GENESIS 4.5...

Considering the number of new subscribers to GA-List (and folks new to
the GA community in general) each week, this question does not surprise
me.  Let me offer a quick history lesson.

Genesis was originally written by John Grefenstette,
and for many years he supplied free copies and updated the 
software as needed.  Sometime ago, Nici Schraudolph and
Rik Belew at UCSD used Grefenstette's Genesis 4.5 as the basis for another
version and this was called GENESIS 1.1ucsd.  This was also freely
distributed. 

John recently updated his system and added a nice user interface for
distribution as part of the Handbook of Genetic Algorithms edited by Dave
Davis.  This is the version available from TSP as described in v5n13.
John no longer distributes the earlier versions.

The UCSD version IS available via anonymous FTP, and it is
described in the next message.  By the way, due to questions like the
above, the authors have changed its name!
 
- Alan C. Schultz

------------------------------

Date: Mon, 10 Jun 91 13:38:38 PDT
From: schraudo@cs.UCSD.EDU (Nici Schraudolph)
Subject: Re:  Genesis ftp site?

   Our GENESIS-based public domain GA package continues to be available by
   anonymous ftp in pub/alife/software/unix/GAucsd/ at iuvax.cs.indiana.edu
   and in pub/GAucsd at sdcsvax.ucsd.edu, as well as by e-mail server (send
   me an empty message with subject "send 1.2ucsd source").

   To avoid confusion with the Grefenstette/Davis diskette as well as the
   abundance of other software packages with similar names, we have decided
   to rename our software "GAucsd".  The package itself remains the same:
   based on GENESIS 4.5, but heavily modified over the years.

   - Nici Schraudolph.

------------------------------
End of Genetic Algorithms Digest
******************************
