
Genetic Algorithms Digest   Thursday, September 12 1991   Volume 5 : Issue 28

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

Today's Topics:
	- Re: order operators
	- Opinions on the proposed GA journal
	- Response: GA Journal Status
	- Scheduling algorithms on parallel hardware

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

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

 First European Conference on Artificial Life (v5n10)         Dec 11-13, 1991
 Canadian AI Conference, Vancouver, (CFP 1/7)                 May 11-15, 1992
 10th National Conference on AI, San Jose, (CFP 1/15)         Jul 12-17, 1992
 ECAI 92, 10th European Conference on AI (v5n13)              Aug  3-7,  1992
 Parallel Problem Solving from Nature, Brussels, (CFP 4/15)   Sep 28-30, 1992

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

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

From: jrv@sdimax2.mitre.org
Date: Fri, 06 Sep 91 18:52:50 EDT
Subject: Re: order operators

   In volume 5, issue 27, I proposed that two crossover operators
   described by Syswerda and Starkweather et al. were equivalent.
   Tim Starkweather replied

   >   After carefully reviewing the operators (along with D. Whitley
   >   and K. Mathias), we agree that what is presented as Order Crossover
   >   #2 and Position-based Crossover in [1] are basically equivalent, if
   >   the number of crossover points averages half the length of the
   >   string.  

   Incidently, it's also equivalent to a uniform order crossover operator
   presented by Lawrence Davis in the tutorial section of his recent book
   [1].  (That may not be his name for the operator.)

   I also suggested that the two operators must have been used with
   different fractions of the genetic material inherited from the two
   parents.  Tim went on to state

   >   This is indeed the reason why the two operators produced different
   >   results...  The number of crossover points used for each operator
   >   was on average less than half the length of the string.

   This suggests that when use of this crossover operator is documented, a
   parameter should be stated: the fraction (say, P_p) of elements
   inheriting absolute position.

   This parameter corresponds to the parameter P_0 in the parameterized
   uniform crossover operator described by Bill Spears and Ken De Jong
   [2].  Note however that only the range 0 < P_0 <= .5 is interesting due
   to symmetry, whereas the full range 0 < P_p < 1 is of interest --
   P_p < .5 emphasizing order and P_p > .5 emphasizing position.

			      - Jim Van Zandt

   [1] Handbook of Genetic Algorithms, L.  Davis, ed., 1990.  

   [2] W. M. Spears and K. A. De Jong, "On the Virtues of
   Parameterized Uniform Crossover,", ICGA-91, pp. 230-236.

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

From: melaniem@zip.eecs.umich.edu
Date:  Sun, 1 Sep 91 13:43:28 EDT
Subject: Opinions on the proposed GA journal

   During the business meeting at ICGA 91, there was a good deal of
   discussion about the possibility of a GA journal.  I don't know what
   "behind-the-scenes" action has taken place since the conference, but the
   moderators at the meeting suggested further discussion on the GA List, so
   I would like to put in my two-cents on this subject.

   I think such a journal might fill an important niche in "journal space",
   but we need to think carefully about what that niche should be, since
   there are already so many other journals and so much overlap among them.
   There is some danger that a purely "genetic algorithms" journal would too
   narrow.  It might tend to exclude work that, while not strictly about
   genetic algorithms, is closely related and very relevant.  Tom Ray's work
   on the Tierra simulator could be an example of this: this work might
   considered by some to be out of the mainstream of GA research.  However, I
   feel that the proposed journal should be structured so that research like
   Ray's on the emergent dynamics of abstract evolution is just as central as
   analysis of GA deception or using GAs to solve job shop problems.  There
   is also a danger of the journal becoming too broad, trying to include
   everything in all these related new fields such as emergent computation,
   artificial life, neural networks, etc.  It is very important that the
   proposed journal have a clear and not overly broad focus, that it not
   overlap too much with existing journals.  Such a lack of clear focus and
   potential for large overlap might be a problem with a journal entitled
   "Natural Computation", as was being proposed at ICGA.

   My opinion is that if we have a journal, its focus should be on the theory
   and applications of computational systems that use ideas from evolution
   (not restricted to "standard" GAs), and on the feedback that research on
   such systems provide to the study of biological evolution, or evolution in
   general.

   A natural name for such a journal would be "Evolutionary Computation"
   (this name was suggested by David Chalmers at the business meeting).  I
   think this is the best name to get across the focus of the journal.  (It
   is also a a nice analogy with the journal "Neural Computation", which
   focuses on the study of computational systems that were inspired by the
   brain, and the feedback this study provides to neuroscience.)

   Topics appropriate for such a journal would be the theoretical analysis of
   such systems, the development of new types of such systems, applications
   of such systems to problems in various disciplines (e.g., function
   optimization, machine learning, evolving neural nets, automatic
   programming), relations between these systems and other search and
   optimization techniques, and implications of theoretical and empirical
   studies of such systems for natural evolutionary systems.

   I don't think this should be a journal for papers about neural networks
   with no evolutionary component, or for papers about other search
   techniques (e.g., simulated annealing) inspired by natural systems.
   Standard computer models of population genetics also would probably not be
   appropriate for this journal, although it would take some thought to
   figure out which topics in population genetics might be appropriate here.
   It also should not be a journal for papers on artificial life that don't
   involve GAs or evolutionary techniques (e.g., computer simulations of ant
   colonies)---there will very likely be a separate Artificial Life journal
   for such papers.  There is, of course, a lot of potential for overlap with
   other journals, and the journal's focus on evolutionary computation should
   be kept clear.

   That's my opinion about this.  Any word from the people organizing the 
   proposed journal on the journal's status right now?

   Melanie Mitchell
   AI Lab
   University of Michigan

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

From: dejong@AIC.NRL.Navy.Mil
Date: Thu, 12 Sep 91 10:26:10 EDT
Subject: GA Journal Status

  I think that Melanie has articulated very well the feelings of many of the
  members of the GA community: the need to carefully define the niche a GA
  journal might fill in a manner which is neither to restrictive nor too
  general.  As early as 1984-5, the organizers of the first GA conference
  discussed these same issues with respect to the name of the conference as
  well as journal possibilities.  Although the immediate issue is one of
  defining and creating a GA journal, there are direct implications for GA
  conferences and the GA society as well.  Let me try to summarize the
  arguments which have been made both for and against various possibilities:

  1. Genetic Algorithms Journal

	  + no other place for "hard core" GA papers
	  + significant name recognition.
	  + natural association with ICGA and ISGA.
	  + increasing activity in GA-specific research.

	  - suggests too narrow a focus.
	  - might discourage other groups interested in other variations
	    of evolutionary algorithms (e.g., ES community).
	  - too small a community to support a journal.
	  - other journals with "algorithms" in the title focus
	    on publishing/discussing code.

  2. Evolutionary Algorithms/Programming/Computation Journal

	  + captures/emphasizes an important shared interest of a broader
	    community (e.g., ES)
	  + encourages development/evaluation/comparison of non-GA 
	    evolutionary systems.
	  + sufficiently large community to support a journal.

	  - no current name recognition or association with existing
	    GA conferences, books, papers, or ISGA.
	  - might discourage groups looking at other biologically motivated
	    computation systems (e.g., immune systems).

  3. Natural/Biological Algorithms/Programming/Computation Journal

	  + captures/emphasizes interests of a broader community.
	  + encourages development/evaluation of non-evolutionary systems.
	  + sufficiently large community to support a journal.

	  - no current name recognition or association with existing
	    GA conferences, books, papers, or ISGA.
	  - too broad: encompasses entire neural net community.
	  - there is already a book and a research group using the
	    phrase "Natural Computation" which focuses on the computational
	    aspects of biological sensory systems.
	  - might discourage groups focusing on optimization and more
	    engineering-oriented problems.


  There has also been considerable discussion concerning the use of the
  terms "algorithms", "programming, "computation", and "systems".  There is
  general agreement that "systems" is too broad a term (e.g., genetic
  systems, evolutionary systems, ...).  It has been argued that having the
  word "algorithms" or "programming" in the title suggests a focus on
  publishing actual code or papers about the process of producing code.
  Thus "computation" is suggested as the best choice.

  So, what's in a name?  Clearly we would like to have a short phrase which
  conveys the sorts of things we want this journal to focus on.  On the
  other hand, it will be an evolving editorial board and editorial policy
  that will determine the kinds of research which will be published in this
  journal, so we shouldn't get too hung up finding the (possibly
  non-existent) optimal pithy phrase.  My sense is that, if we took a vote
  right now, the three leading candidates would be (in rank order):

	  1. Evolutionary Computation
	  2. Genetic Algorithms
	  3. Natural Computation


  As far as the discussions with potential publishers goes, there
  are currently 3 who have expressed strong interest:

	  1. Elsevier (publishes AI Journal)
	  2. Kluwer (publishes Machine Learning Journal)
	  3. IEEE Publications

  All have expressed the feeling that the scope of the journal should
  probably be broader that just GAs in order to appeal to a large
  enough community to support the journal.  Other than that, they
  have left the decision of title and scope to us.  There are enough
  senior researchers in the field who are willing to serve on the
  editorial board of such a journal.  What remains at this point
  is to negotiate an agreement with one of the publishers, hopefully
  during the next 1-2 months.  Appearance of the first issue would
  be 6-12 months after that.

  All of this, of course, is intended to provide our research community
  with a journal which is useful and appropriate.  So, your comments to 
  me and/or the GA community as a whole concerning this whole process 
  are always welcome.

  Kenneth De Jong

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

From: Andrew Nisbet (PhD) <andy@spec0.electrical-engineering.manchester.ac.uk>
Date: Tue, 10 Sep 91 11:15:18 BST
Subject: Scheduling algorithms on parallel hardware

   My research project involves the optimal scheduling of DSP algorithms onto
   arbitrary parallel hardware structures. The DSP algorithms are represented
   as signal flow graphs and the proposed target hardware is also represented
   graphically.

   I am interested in using Simulated Annealing and Genetic Algorithms in
   comparison with standard heuristic methods. If anyone has some pointers as
   to some books, algorithms or references which may be helpful could they
   please Email me at the address below.

   Andy Nisbet, Dept. of Electrical Engineering, University of Manchester,
		   Manchester M13 9PL, England.
   Internet: andy@spec0.ee.man.ac.uk         Janet: andy@uk.ac.man.ee.spec0
   ARPA: andy%ee.man.ac.uk@nsfnet-relay.ac.uk Wet String: (+44)-61-275-4561

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