
Genetic Algorithms Digest   Wednesday, June 8, 1994   Volume 8 : Issue 19

 - Send submissions to GA-List@AIC.NRL.NAVY.MIL
 - Send administrative requests to GA-List-Request@AIC.NRL.NAVY.MIL
 - anonymous ftp archive: FTP.AIC.NRL.NAVY.MIL (Info in /pub/galist/FTP)

Today's Topics:
	- Re: One offspring or two? (3 messages) (v8n18)
	- About exponential Growth of above average schemata
	- SAB94 Program and Registration
	- ASI-AA-95: Practice and Future of Autonomous Agents
	- TECOM AI Conference - Call for Papers
	- A distributed implementation of GENESIS
	- GA in discrete event simulation

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

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

The IEEE Conference on Evolutionary Computation, Orlando(v7n26) Jun 26-30, 94
FOGA94 Foundations of GAs Wkshop, Estes Park, Colorado(v7n26)Jul 30-Aug 3, 94
SAB94 3rd Intl Conf on Sim of Adaptive Behavior, Brighton(v7n11) Aug 8-12, 94
ECAI-94, 11th European Conference on AI, Amsterdam (v7n23)       Aug 8-12, 94
ECAI-94 Wkshp on Applied Genetic & Other Evol Algs, Amsterdam(v8n5) Aug 9, 94
IEEE/Nagoya Univ WW Wkshp on Fuzzy Logic & NNs/GAs, Japan(v7n33) Aug 9-10, 94
ISRAM94 Special Session on Robotics & GAs, Maui, Hawaii (v7n22) Aug 14-17, 94
TECOM AI Conference (v8n19)                                     Sep 13-16, 94
Evolution Artificielle 94, Toulouse, France (v8n10)             Sep 19-23, 94
COMPLEX94 2nd Australian National Conference, Australia (v7n34) Sep 26-28, 94
PPSN-94 Parallel Problem Solving from Nature, Israel (v7n32)     Oct 9-14, 94
GAs in Image Processing and Vision Colloquium, Savoy Place (v8n16) Oct 20, 94
AI'94 Workshop on Evol Comp, Armidale, NSW, Australia (v8n15)      Nov 22, 94
EP95 4th Ann Conf on Evolutionary Programming, San Diego,CA(v8n6) Mar 1-4, 95
ICANNGA95 Intl Conf on Artificial NNs and GAs, France (v8n10)   Apr 18-21, 95
ECAL95 3rd European Conf on Artificial Life, Granada, Spain(v8n5) Jun 4-6, 95
ASI-AA-95 Practice and Future of Autonomous Agents (v8n19) Sep 23 - Oct 1, 95

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

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

From: Christopher Cilley <cdcilley@oxytricha.mit.edu>
Date: Thu, 02 Jun 94 19:30:08 -0400
Subject: Re- One offspring or two? (v8n18)

In v8n18, Nick Radcliffe <njr@epcc.ed.ac.uk> asks:
>Does anyone have any good reasons for producing two children, rather
>than just one, from crossover?
>...
>Anyway, what do other people think?   I'd be interested if anyone has
>any good reasons for using producing two children (or more, or indeed,
>any other reasons for producing one.)

We have started looking at using GAs to generate structures for proteins
and polynucleotides that satisfy a variety of constraints.  A single
chromosome represents one candidate structure, the genes within it
are numbers that represent the torsion angles within the molecule.

3-D structures of biological macromolecules can have definite
domains within the overall structure.  Therefore, while one part
of the chromosome may code for a fold that satisfies constraints
relevant to that region of the molecule, another part may be way off.

If two such parents are chosen for crossover, and two reciprocal children
are created, there is no way to tell which child is getting the "good"
parts from both parents.  Therefore, we allow both children back into the
population and let their fitnesses determine their fate.


	Christopher Cilley
	cdcilley@oxytricha.mit.edu

"If people were more concerned with being reconciled than
 with being right, the world would be a better place."
		-Judith Martin (aka Miss Manners)

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

From: Peter Ross <peter@aisb.edinburgh.ac.uk>
Date: Fri, 3 Jun 94 09:52:51 BST
Subject: Re: More than one child ... (v8n18)

Nick, concerning your question about producing more than one child
in v8n18...

My tuppence worth: consider chromosomes as denoting points on a
hypercube. If we're talking binary chromosomes then the points are
corners; if not, interior grid points (if alleles are drawn from some
finite set). Producing "twins" means that you produce diagonally
opposite points in/on the hypercube defined by having the parents at
opposing corners. This avoids the accidental bias introduced by taking
just one child of lower-fitness, or even of higher-fitness; it will be
up to selection to apply that bias.

It seems as though producing just one child ought to be
equivalent, since the other child can also be produced by the same
mechanism. But producing the one child FIRST biases the chance of
ever producing the other; indeed, the parents might get elbowed out
by the results of that first child if it is of high fitness. That
might be a pity if the parents are in fact repositories of important
diversity that will be needed later to avoid some non-global maximum.

(This argument might be sharpenable, especially if the definition of
the fitness function could be continued from just the grid points to
the whole of the space. Once you've got a continuous function life
is pretty rosy...)

Some recent experiments I've being doing with pga (version 2.6 is about
to appear) seem to bear this out, though just in terms of comparing
performance with/without twins. I haven't yet delved into tracking
individual genes. Could be a nice little student project in there
somewhere...

Peter Ross
Department of AI
Edinburgh

[ WMS - my thoughts on this topic are similar to those above.  It appears
  to me as if producing two children will yield greater exploration (when
  using bits), which is often useful in a GA. However, this is only a
  conjecture! ]

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

From: wren@scs.leeds.ac.uk
Date: Mon, 6 Jun 94 10:12:27 BST
Subject: Re: More than one child... (v8n18)

Nick Radcliffe's question as to whether two children are better than one is
complementary to one I should like to raise:  are three parents better than
two?   Using what I guess Nick would call an allelic representation we have
produced children which are essentially selections of values drawn from the
parents.   We have some evidence that drawing from three parents in our
particular problem domain produces better offspring than drawing from two
parents.   Experiments with four and five parents produced worse children.

A paper describing this work will appear in Computers and Operations Research:

   A. Wren and D.O. Wren.   A genetic algorithm for public transport driver
   scheduling.

Returning to Nick's question, we chose only one child (without much thought as
to reasons - it just seemed natural, and we had not read anything about GA's
when we started other than a short article by Nick in New Scientist).   It now
seems to me that selecting a single child allows greater diversity because
the genes of more pairs of parents are represented in the next generation, and
this should be a good thing.   This is certainly the case with an allelic
representation.   However, a binary representation with standard crossover
yields two children which are very different.   Selecting both probably
maintains diversity and must give a slight increase in speed in cases where the
selection of pairs of parents may be complex.

           Tony Wren
           Director, Scheduling and Constraint Management Group
           School of Computer Studies
           University of Leeds
           Leeds LS2 9JT
           UK
           email:   wren@uk.ac.leeds.scs
           phone:   (44) (0)532 335481
           FAX:     (44) (0)532 335468

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

From: carmona@madres.cert.fr (Olivier Carmona)
Date: Tue, 31 May 94 16:09:20 +0200
Subject: About exponential Growth of above average schemata.

      In "Genetic Algorithms in Search, Optimization, and Machine Learning"
      by D.E. Golberg, I read p 30 :
      "...
       The effect of reproduction on the number of schemata is qualitatively
clear; above-average schemata grow and below-average schemata die off. Can we
learn anything else about the mathematical form of this growth from the schema
difference equation ? Suppose we assume that a particular schema H remains above
                   _
average an amount cf with c a constant. Under this assumption we can rewrite the
schema difference equations as follows :
                          _    _
                          f + cf
       m(H,t+1) = m(H+t) _______ = (1 + c) . m(H,t)
                            _
                            f
       Starting at t = 0 and assuming a stationary value of c,..."

       The conclusion of this paragraph is the exponentially increasing
allocation of trials for above average schemata.
       But the assumption that c has a stationary value come with no proof, is
there someone that can give me a proof of this.

       Olivier Carmona

[ WMS: Ken De Jong and Jayshree Sarma address this issue in their paper,
  "Generation Gaps Revisited", in the proceedings of the 1992 Foundations
  of Genetic Algorithms workshop. ]

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

From: "Phil Husbands" <philh@cogs.susx.ac.uk>
Date: Fri, 20 May 1994 11:58:08 +0100 (BST)
Subject: SAB94 Program and Registration

               CONFERENCE PROGRAM AND INVITATION TO PARTICIPATE
               ------------------------------------------------


                            FROM ANIMALS TO ANIMATS
     Third International Conference on Simulation of Adaptive Behavior (SAB94)


                        Brighton, UK, August 8-12, 1994


The object of  the conference is  to bring together  researchers in  ethology,
psychology,  ecology,  cybernetics,  artificial  intelligence,  robotics,  and
related fields  so  as to  further  our  understanding of  the  behaviors  and
underlying mechanisms that allow animals and, potentially, robots to adapt and
survive in uncertain environments.

The conference  will  focus  particularly  on  well-defined  models,  computer
simulations, and  built  robots in  order  to help  characterize  and  compare
various  organizational  principles  or  architectures  capable  of   inducing
adaptive behavior in real or artificial animals.

[ WMS: The full program, registration material, etc., can be found on
  our FTP site under /pub/galist/info/conferences/SAB94 ]

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

From: Nikolaus Almassy <almassy@ifi.unizh.ch>
Date: Sat, 21 May 94 11:16:19 +0200
Subject: ASI-AA-95: Practice and Future of Autonomous Agents

  First announcement:

                Practice and Future of Autonomous Agents:
			       ASI-AA-95

		 23  September  -  1st  October  1995
		      Centro  Stefano  Franscini
		 Monte  Verita,  Ticino,  Switzerland

  (Follow-up meeting of the NATO Advanced Study Institute "The Biology and
Technology of Intelligent Autonomous Agents", which took place Spring 1993
in Trento, Italy).

Further information

Please contact:  Rolf Pfeifer
AI Lab, Computer Science Department
University of Zurich, Switzerland
E-mail:  pfeifer@ifi.unizh.ch
Fax:  +41-1-363 00 35
Phone:  +41-1-257 43 20/31

Updates of this announcement

can be retrieved by
ftp://claude.ifi.unizh.ch/pub/monteverita/ASI-AA-95.txt or by sending an
empty E-mail to <info@ifi.unizh.ch> with "monte verita" in the subject
line.  It can also be accessed by World Wide Web at the the URL:
http://josef.ifi.unizh.ch/asi-aa.html

Please distribute this announcement to those who might be interested.

Nikolaus Almassy
Department of Computer Science  almassy@ifi.unizh.ch
University of Zurich            Tel: +41 - 1 - 257 43 25
8057 Zurich, Switzerland        Fax: +41 - 1 - 363 00 35

[ WMS: The full announcement can be found on our FTP site under
  /pub/galist/info/conferences/ASI-AA-95 ]

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

From: Jeffrey Murter (STECS-DAI 3711) <jmurter@APG-9.APG.ARMY.MIL>
Date:     Thu, 26 May 94 8:46:53 EDT
Subject:  TECOM AI Conference - Call for Papers

TECOM Artificial Intelligence Technology Symposium

Theme: AI Technology from the Laboratory into the Workplace

Edgewood Conference Center Edgewood Area
Aberdeen Proving Ground, Maryland

Critical Dates:
Receipt of Abstracts:  June 15, 1994
Author Notification:   June 30, 1994
Preregistration Deadline:   September 2, 1994
Symposium Dates:  September 13-16, 1994

[ WMS: The full announcement can be found on our FTP site under
  /pub/galist/info/conferences/TECOM ]

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

From: Erick Cantu Paz <ecantu@lamport.rhon.itam.mx>
Date: Thu, 26 May 1994 12:33:15 -0600 (CST)
Subject: A distributed implementation of GENESIS

I have finished a stable version of Distributed GENESIS (DGENESIS). It is
based in John Grefenstette's GENESIS 5.0 and it lets the user set the number
of subpopulations, the migration rate, the migration interval and the
migration topology. There have not been many studies on the effect of the
topology on the performance of a PGA, and I think DGENESIS can encourage
experimentation in this area.

DGENESIS still needs many refinements, but I think someone might find it
useful. It is available by anonymous ftp from lamport.rhon.itam.mx and
ftp.aic.nrl.navy.mil.

Erick Cantu-Paz
Instituto Tecnologico Autonomo de Mexico (ITAM)
email: ecantu@lamport.rhon.itam.mx

[ WMS: The code is available on ftp.aic.nrl.navy.mil under
  /pub/galist/src/ga/dgenesis-1.0.tar.Z. Anyone willing to do
  a review of this software? ]

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

From: msalz@ws1.atv.tuwien.ac.at (Manfred Salzmann)
Date: Sat, 28 May 1994 14:53:53 +0200 (MET DST)
Subject: GA in discrete event simulation

Dear GA-digest-readers,

I'm doing my PhD-thesis about optimization in discrete event simulation
using GAs. To evaluate my fitness-function I want to use a simulation
language called GPSS/H, but it seems that noone had done such an
application using GPSS/H and therefore I want to ask if there is
somebody out there who can provide me with some additional information on
this special topic.
Any help will be appriciated (please email to msalz@ws1.atv.tuwien.ac.at)!

	nice greetings from Vienna
		M. Salzmann

P.S.: If anybody of you is interested in Simulation: Next year the European
Simulation Congress (which takes place every third year) will be held in Vienna.
For further information please email to eurosim95@email.tuwien.ac.at .
<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
Manfred Salzmann	Vienna University of Technology		Dept. 114/5
Wiedner Hauptstr. 8-10		A-1040 Vienna		Austria (Europe)
			      (+43-1) 58801 5419
msalz@ws1.atv.tuwien.ac.at 			manfredo@simserv.tuwien.ac.at
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>

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

