
Genetic Algorithms Digest   Monday, April 26, 1993   Volume 7 : Issue 9

 - 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:
	- GA with Sharing Details
	- applications in economics
	- references on simulated annealing
	- N.Net & G.A software
	- paper available
	- PPSN-94 Announcement
	- Genetic Programming Software - 2 messages
	- Adaptive Simulated Annealing (ASA)

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

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

ECAL-93, 2nd European Conference on A-Life, Brussels (v6n31)    May 24-26, 93
CSCS93, 9th Int Conf on control systems & CS, Romania (v7n3)    May 24-27, 93
ANN93, IEE Intl Conf on Artificial Neural Nets, Brighton        May 25-27, 93
ICGA-93, Fifth Intl. Conf. on GAs, Urbana-Champaign (v6n29)     Jul 17-22, 93
COLT93, ACM Conf on Computational Learning Theory, UCSC (v6n34) Jul 26-28, 93
Machine Learning & Knowledge Acq. Workshop (IJCAI), France (v7n1)  Aug 29, 93
IEE/IEEE Workshop on Nat Alg in Signal Processing, Essex (v7n5) Nov 15-16, 93
EP94 3rd Ann Conf on Evolutionary Programming, San Diego (v7n7) Feb 24-25, 94
ISEC-94 Int. Symp. on Evolutionary Computation, Orlando (v6n40) Jun 25-30, 94
PPSN-94 Parallel Problem Solving from Nature, Israel (v7n9)      Oct 9-14, 94

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

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

From: chank@ecn.purdue.edu (King Chan)
Date: Mon, 29 Mar 93 18:25:45 -0500
Subject: GA with Sharing Details

   Greetings,

   I have a few questions on the details of using GA's with Sharing
   functions [Goldberg and Richardson] as the exact implementation
   is not very clear to me.

   The proposed sharing method is supposed to enable GA's to maintain
   stable subpopulations in multiple regions of the solution space by
   derating regions in which a niche has formed.  In this manner other
   regions may dominate and be explored by the GA.

   My current understanding of GA's with sharing is illustrated in fig 1


			__________________
    |------------> |--->|   Standard GA   |
    |              |    ------------------
    |           No |            |
    |              |            V
    |              |     _____________________
    |              <-   /  Performance High   \  Measure via some performance
    |                   \_____________________/  measure. i.e., Offline perf.
    |                            |
    |                            V  Yes
    |                    _________________________
    |                --->| Apply Sharing Function|
    |               |    -------------------------
    |               |            |
    |               |            V
    |             No|     _____________________
    |               <-   /  Has Niche Formed ? \  Measure via some performance
    |                    \_____________________/  measure. i.e., Offline perf.
    |                            |
    |                            V
    |                    __________________
    |                    | Save Solution  |
    |                    ------------------
    |                           |
     - To Standard GA    <------ V


                        Figure 1.  GA with Sharing Flowsheet


   [1] D.E. Goldberg and J. Richardson, "Genetic algorithms with sharing
   for multimodal function optimization", 2nd ICGA (1987).

   Thus, one only enters sharing when the solution is very close to
   forming a niche.  This is currently measured by the offline performance.
   Sharing is exited once a niche has formed.  One or more niche
   solutions are then saved and may participate in the reproduction
   process but with the derated fitness.

   My question are as follows:

   o  Is this interpretation of the application of sharing functions correct ?

   o  Where exactly is diversity introduced since the derated fitness members
      may still be the dominate members in the population and again be sampled
      in the standard GA.  This is exactly the case for multimodal functions of
      varying depth with one extra deep well ?

   o  I guess I would like to know your versions of figure 1.

   Thanks very much.

   chan@ecn.purdue.edu

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

From: "Roger A. McCain" <MCCAINRA%DUVM.BITNET@pucc.Princeton.EDU>
Date: Fri, 05 Mar 93 08:55:59 EST
Subject: applications in economics

   I have done some pretty preliminary work on applications of GA-like
   evolving heuristics in economics. I think I came in too late for the
   call for economics work, but I also have a comment and a question for
   the general community.  My models simulate the emergence of demand
   heuristics. (They are limited to the demand side only, unlike the
   work on auction markets). Mostly, they have been applied to the
   economics of the arts, where the interesting complication is that
   the consumer must learn to appreciate the good. (Anyway, that fit into
   the program for the conference in Venice!)
   COMMENT I had begun with some non-GA-like, rather Nelson-and-Winter-like
   simulations that I called "Simple Groping." They gave rise to quite
   "inelastic" demand heuristics, i.e. nonresponsive to price changes.
   I tried using GA's to code for more responsive demand heuristics,
   with some success. However, the GA's lack a property that groping
   (and Nelson-and-Winter) models have. I call the property
   "teleological conservatism" and will call the algorithms TCA's for
   the obvious sort of reason. Like this: in a GA, as in organic evolution,
   the filial population replaces the parent population despite the fact
   that some offspring may be less adapted than the parents they replace.
   In a TCA, however, the offspring replace the parent only when they are
   better adapted. One result seems to be (on very limited experiments)
   that TCAs do not seek the global optimum as GA's do, but instead
   distribute simulated agents among the local optima. Thus, TCA's would
   be less effective in optimizing certain functions than GA's, but might
   make more sense as simulations of real human heuristic processes.
   QUESTION While my simulations are artificially simplified, so far as
   their optimization aspects are concerned, and converge only to
   approximately optimal heuristics, they do seem to converge fairly
   fast. I note that many of the applications are allowed to run for
   millions of iterations. Mine run for several digits less. Am I doing
   something wrong? Here is one possibility: I don't get these results
   unless the initial conditions are tinkered a bit. I can do that, of
   course, because with simulations I have the optima to begin with. When
   the initial conditions are far off, the simulated populations lose
   the diversity very fast and converge to very bad results that persist
   for hundreds of iterations. Perhaps this would be reversed in millions
   of iterations -- or have I missed something fundamental here? Are
   there any results on rate of convergence that might speak to this?

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

From: suchi@pollux.cs.uga.edu (Suchi Bhandarkar)
Date: Thu, 1 Apr 1993 09:33:21 -0500 (EST)
Subject: references on simulated annealing

   Could anybody mail me a list of references on parallel
   implementations of simulated annealing algorithms. I am
   particularly interested in implementations on SIMD machines
   such as CM-2 or the MasPar MP computers. Kindly e-mail your responses
   to "suchi@pollux.cs.uga.edu" Thank you very much.

					----- Suchi Bhandarkar

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

From: David Bradbury <D.C.Bradbury@open.ac.uk>
Date: 6 Apr 93 15:27:04 U
Subject: N.Net & G.A software

   does anyone know where I can get hold of public domain software to simulate
   neural networks and/or generic algorithms. are there any packages which
   allow one to combine these two procedures?

   Thanks!

   David

   d.c.bradbury@open.ac.uk

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

Date: Wed, 17 Mar 93 15:53:31 MST
From: mm@santafe.edu
Subject: paper available

  The following paper is available by public ftp.

			Revisiting the Edge of Chaos: 
	      Evolving Cellular Automata to Perform Computations

     Melanie Mitchell       Peter T. Hraber         James P. Crutchfield
    Santa Fe Institute    Santa Fe Institute  University of California,Berkeley

		Santa Fe Institute Working Paper 93-03-014

				  Abstract

  We present results from an experiment similar to one performed by
  Packard (1988), in which a genetic algorithm is used to evolve 
  cellular automata (CA) to perform a particular computational task.  Packard
  examined the frequency of evolved CA rules as a function of Langton's 
  lambda parameter (Langton, 1990), and interpreted the results of his 
  experiment as giving evidence for the following two hypotheses:
  (1) CA rules able to perform complex computations are most likely
  to be found near ``critical'' lambda values, which have been claimed
  to correlate with a phase transition between ordered and chaotic behavioral
  regimes for CA;  (2) When CA rules are evolved to perform a complex
  computation, evolution will tend to select rules with lambda values
  close to the critical values.   Our experiment produced very different
  results, and we suggest that the interpretation of the original results
  is not correct.  We also review and discuss issues related to lambda, 
  dynamical-behavior classes, and computation in CA.  

  The main constructive results of our study are identifying the emergence 
  and competition of computational strategies and analyzing the central
  role of symmetries in an evolutionary system. In particular, we
  demonstrate how symmetry breaking can impede the evolution toward
  higher computational capability.


  To obtain an electronic copy:

	  ftp santafe.edu
	  login: anonymous
	  password: <your email address>
	  cd /pub/Users/mm
	  binary
	  get rev-edge.ps.Z 
	  quit

  Then at your system:

	  uncompress rev-edge.ps.Z
	  lpr -P<printer-name> rev-edge.ps


  To obtain a hard copy, send a request to mm@santafe.edu.  

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

From: yuval@weizmann.ac.il (Davidor Yuval)
Date: Tue, 13 Apr 93 11:36:28 +0300
Subject: PPSN-94

  Dear colleague,

  I am happy to announce the details of the next PPSN conference which
  will be held in Israel between 9-14 October 1994.                                               
  I would like to point out the danger in defusing the central 
  role of the ICGA and PPSN conferences.  Recently a symposium on evolutionary
  computation was announced to be held in Orlando, Florida, between June 25 -
  July 1, 1994, dates which are very close to the traditional dates of the PPSN
  conferences.  Of course there is no restriction on individuals wishing to
  organize a conference on evolutionary computation, but it is counter
  productive for the community as a whole to have too many competing channels
  for international exchange of ideas and publications.  It was agreed by the
  American and European communities that in alternate years an ICGA and a PPSN
  international conference will be held, and that these are the official
  international conferences of the community. I hope that members of the
  community will demonstrate their support of these two conferences by giving 
  priority for submission, and  attending the ICGA and PPSN conferences.

  Looking forward in welcoming you in Israel,

  Yuval Davidor.

		      The third international conference on 
		      Parallel Problem Solving from Nature
				  (PPSN III)

			       First Announcement
				      and
				 Call for Papers

  The third international conference on Parallel Problem Solving from Nature 
  (PPSN III) will be held in Jerusalem, Israel between 9-14th October, 1994. 
  This meeting will bring together an international community from academia, 
  government and industry interested in algorithms suggested by the unifying 
  theme of natural computation.

  Natural computation is a common name for the design, theoretical and
  empirical understanding of algorithms gleaned from nature. Characteristic 
  for natural computation is the metaphorical use of concepts, principles and 
  mechanisms underlying natural systems.  Examples are genetic algorithms,
  evolutionary programming and evolution strategies inspired by the 
  evolutionary processes of mutation, recombination, and natural selection 
  in biology, simulated annealing inspired by many--particle systems 
  in physics, and algorithms inspired by multi--cellular systems like neural 
  and immune networks.

  Topics of particular interest include, but are not limited to: 
	  evolution strategies, evolutionary programming, 
	  genetic algorithms and classifier systems, 
	  other forms of evolutionary computation, 
	  simulated annealing, neural and immune networks, 
	  machine learning and optimization using these methods, 
	  their relations to other learning paradigms, 
	  and mathematical description of their behaviour.  

  An emphasized theme of the conference will be the successful application 
  of these techniques to solve real problems in manufacturing, design, 
  planning and engineering.

  The conference will be held in a kibbutz 10 min. from the old city of
  Jerusalem on top of the Judea mountain ridge overlooking Bet-Lehem and
  Jerusalem.  The conference programme will include visits to historical,
  religious and contemporary monuments in Israel.

  Important Dates:

  1  February   1994: Submissions due
  30 April      1994: Notification to authors
  1  June       1994: Revised, final camera-ready papers due
  9--14 October 1994: Conference dates

  PPSN-94 Conference committee:

  Conference Chair: 	   Y. Davidor (Israel)

  Programme Co-Chairpersons: H.-P. Schwefel (Germany)
			     and      
			     R. Maenner (Germany)

  Official Organizer & travel Arrangements: 
			     Ortra Ltd., 2 Kaufman St.,
			     POBox 50432, Tel-Aviv 61500, Israel.
			     Tel. :  +972-3-66.48.25
			     FAX  :  +972-3-510.21.98

  For questions regarding the programme, please contact one of the programme
  co-chairs.  For any traveling or organizational query please contact
  Mr. Daniel Tidar of Ortra stating clearly the conference name and date.

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

From: Howard@quercus.demon.co.uk
Date: Sat, 6 Mar 1993 19:53:45 +0000
Subject: New upload - Genetic Programming Software

   I have put the following file in [src/ga/koza.gp]
   [Ed's Note: This has been modified to reflect the current ftp pathname.
    See the next message for full ftp information.  -- Connie]
   being in (Unix format) ASCII text.

   This code was written by John Koza <koza@cs.stanford.edu> and James Rice
   <rice@sumex-aim.stanford.edu>, and is being posted by me with their full
   consent (I am acting on behalf of the gp mail list mentioned below, and
   loosely controlled by James Rice).

   The following is info for the README:

   koza.gp is a pure (CLtL2) Common Lisp implementation of the Genetic
   Programming Paradigm, as described in "Genetic Programming" by John R Koza,
   MIT Press, 1992 (ISBN 0-262-11170-5, MIT Press order code KOZGH).  Great
   care has been taken to ensure that the code both works and is identical to
   the code shown in the book.  The exception to this is the set of
   "top-level" forms that are shown in the appendix as examples, for instance,
   of calls that would fire off the GPP.  These have not been included in-line
   in the code so as to prevent execution of the system during the
   compile/load cycle.  All of these test expressions haven't been included in
   one test function.

   This code is Copyright (c) John Koza, All rights reserved.
   U.S. Patent #4,935,877.  Other patents pending.

   Any enquiries relating to this code, or to genetic programming in general,
   should be directed to the GP discussion group mailing list.  The list is
   called genetic-programming@cs.stanford.edu
   To subscribe to it, you should send mail to
       genetic-programming-request@cs.stanford.edu 
			  ^^^^^^^^ 
   clearly giving the address you with to have mail sent to.

   Howard.
   Howard Oakley,                      * Howard@quercus.demon.co.uk
   EHN & DIJ Oakley, Brooklands Lodge, * AppleLink UK0392
   Park View Close, Wroxall, Ventnor,  * CompuServe 70734,120
   Isle of Wight UK PO38 3EQ           * Tel +44 983 853605, fax 853253 

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

From: schultz@AIC.NRL.Navy.Mil
Date: Thu, 15 Apr 93 09:26:02 EDT
Subject: Genetic Programming Software added to ga-list archive site

   I have added the following submitted software to the ga-list ftp server.
   The ftp machine's address is: ftp.aic.nrl.navy.mil (192.26.18.74)
   The path name for this new package is: /pub/galist/src/ga/koza.gp

   New Package   Description
   -----------   -----------

   koza.gp	 Koza.gp is a pure (CLtL2) Common Lisp implementation of the
		 Genetic Programming Paradigm, as described in "Genetic
		 Programming" by John R Koza, MIT Press, 1992 (ISBN
		 0-262-11170-5).  Written by John Koza <koza@cs.stanford.edu>
		 and James Rice <rice@sumex-aim.stanford.edu>.  This code is
		 Copyright (c) John Koza, All rights reserved.  U.S. Patent
		 #4,935,877.  Other patents pending.  Submitted by Howard
		 Oakley (Howard@quercus.demon.co.uk).

   --Alan C. Schultz

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

From: Lester Ingber <ingber@alumni.cco.caltech.edu>
Date: Fri, 23 Apr 1993 12:09:20 -0700
Subject: Adaptive Simulated Annealing (ASA)

Adaptive Simulated Annealing (ASA)

To get on or off blind-copy ASA e-mailings, just send an e-mail to
ingber@alumni.caltech.edu with your request.

I have updated the Netlib and Statlib Very Fast Simulated Reannealing
(VFSR) code, now at version 9.4, to this release of ASA.  The code
has substantially evolved since its first form in 1987 through its
public release in Nov 92, and this name change reflects this.

                NETLIB (compressed share file)
Interactive:
        ftp research.att.com
        [login as netlib, your_login_name as password]
        cd opt
        binary
        get asa.Z
Email:
	mail netlib@research.att.com [AT&T Bell Labs, NJ, USA]
	mail netlib@cs.uow.edu.au [U Wollongong, NSW, Australia]
and send the one-line message
        send asa from opt
(It may take a week or so for the code in research.att.com to propagate
to the other netlib sites.)

                STATLIB (uncompressed share file)
Interactive:
        ftp lib.stat.cmu.edu
        [login as statlib, your_login_name as password]
        cd general
        get asan
Email:
        mail statlib@lib.stat.cmu.edu
and send the one-line message
        send asan from general

If you do not have ftp access, get information on the FTPmail service
by sending the word "help" as a message to ftpmail@decwrl.dec.com.
If you receive ASA via e-mail, then first `uudecode mailfile',
(where mailfile may be a synthesis of several files) to get asa.Z,
and then follow the previous directions.

If this is not convenient, and if your mailer can handle large files,
I directly can send you the code or papers you require via e-mail.
(I have placed a file ingber.tar.Z of papers in ftp.uu.net:/tmp which
can be retrieved via anonymous ftp.)  Sorry, I cannot assume the task
of mailing out hardcopies of code or papers.

Lester
========================================================================

|| Prof. Lester Ingber                                               ||
|| Lester Ingber Research                                            ||
|| P.O. Box 857                                                      ||
|| McLean, VA  22101                EMail: ingber@alumni.caltech.edu ||

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

