
Genetic Algorithms Digest    Wednesday, 22 November 1989    Volume 3 : Issue 19

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

Today's Topics:
	-  Pittsburgh GA Interest Group
	-  pattern learning plan
	-  Re: GA and NNs (and other ML architectures)
	-   GA in AI?
	-  Call for Abstracts


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

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

1-2 Dec 89 - NIPS Workshop: NNs and GAs (v3n16)
15-19 Jan 90 - IJCNN Session on Evolutionary Processes (v3n10)
Mar 90 - Double Auction Tournament - Sante Fe Institute  (v3n12)
9 May 90 - Workshop on GAs, Sim. Anneal., Neural Nets - Glasgow (v3n15)
21-23 Jun 90 - 7th Intl. Conference on Machine Learning (submissions 1 Feb 90)
15-18 Jul 90 - Workshop Foundations of GAs (THIS ISSUE)
(Send other activities to GA-List@aic.nrl.navy.mil)

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

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

Date: Tue, 14 Nov 89 09:24:45 EST
From: Clay Bridges <Clayton.Bridges@A.GP.CS.CMU.EDU>
Subject: Pittsburgh GA Interest Group


                            Pittsburgh 
               Genetic Algorithms Interest Group (GIG)

We have formed an informal interest group for genetic algorithms,
classifier systems, and related subjects. We maintain a mailing list,
and presently meet about once a month. If you are going to be in
Pittsburgh, and you have an interest in talking with group members, or
in giving an informal talk to the group or a seminar arranged by the
group, please let me know.

Clay Bridges
Graduate Student, Carnegie Mellon CS
clay@cs.cmu.edu
(412) 268-3043

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

Date: Wed, 15 Nov 89 15:00:45 -0500
From: androula@helium.ecn.purdue.edu (Ioannis Androulakis)
Subject: pattern learning plan


   I am interested in getting information concering
   the pattern learning plan, so that patterns of
   changing that cause improvement can be explored
   from one generation to another.
   Further, I would like once again to ask whether        
   any work concerning the time bound for GAs to
   converge, has been done.
   Thank you,
   Ioannis P. Androulakis

   e-mail : androula@helium.en.purdue.edu

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

Date: Thu, 16 Nov 89 22:21:00 -0600
From: honavar@cs.wisc.edu (Vasant Honavar)
Subject: Re: GA and NNs (and other ML architectures)

I agree with the sentiments expressed by Hammerstrom and Dietterich
that the use of evolutionary learning methods will probably not speed 
up the training time required on problems like NETTALK. 

However, evolutionary learning techniques seem to offer a whole range of
possibilities for automated (evolutionary) design of complex network 
topologies / knowledge representation schemes / mix of different machine 
learning paradigms .... to solve complex problems or sets of tasks by an
integrated system which cannot be solved by the use of single prespecified 
network topology, a single machine learning technique, or a single knowledge 
representation scheme. As computing power continues to become cheaper by the
day, entire networks of workstations can serve as the environment in which
different evolutionary variants can continue to learn, compete, and reproduce,
and perish. Most of the workstations are idle a good fraction of the time 
anyway. We can better utilize the otherwise wasted computing resources of idle 
workstations to carry out such large scale evolutionary-machine-learning
experiments. Software environments are already being developed to seek out
and utilize the computing resources of microvaxes on a network (e.g.,
CONDOR - program developed at University of Wisconsin, Madison).

Vasant Honavar
Computer Sciences Dept.
University of Wisconsin
1210 W. Dayton St.
Madison, WI 53706.
honavar@cs.wisc.edu

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

Date:     Fri, 17 Nov 89 13:30:40 GMT
From: pat%computer-science.strathclyde.ac.uk@NSFnet-Relay.AC.UK
Subject:  GA in AI?

Does GA belong in AI?

I am currently putting together a 4th year AI course (here at the
University of Strathclyde). As part of the course I intend to cover
the more conventional search mechanisms. I am planning, later on in
the course, at least 3 lectures on (a) GA, (b) SA, (c) SIGH. I can
apply these techniques to problems addressed in earlier lectures (such
as TSP, bin packing, scheduling, Ackley's functions).

One of my colleagues reckons this is the wrong forum for GA, SA,
SIGH. 

My personal view
 I think its important that the students get to know about this
 work. If I dont give these lectures ...... who will?

I do not want to start the scenario we see in AI-List (ie free will
etc). I'd like a reply from just a few, then kill it dead, and get on
with the work!

Over and out.

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

Date: Mon, 20 Nov 89 18:15:34 -0500
From: Gregory J. E. Rawlins <rawlins@iuvax.cs.indiana.edu>
Subject: Call for Abstracts


                             Call for Abstracts

                               A Workshop on
        The Foundations of Genetic Algorithms and Classifier Systems

                            July 15th-18th, 1990
                         Bloomington, Indiana, USA



	The main thrust of this workshop will be towards rigorous approaches
	to understanding the foundations of genetic algorithms and classifier
	systems. Experimental results are acceptable, but they should have a
	defensibly (and preferably explicit) theoretical basis.

	Acceptable topics include but are not limited to:
	 - convergence properties of genetic algorithms
	 - structure theory for GA-hard problems
	 - properties of the schema theorem
	 - coding functions and representation
	 - genetic algorithms as combinatorial optimizers
	 - parallelizing evolutionary systems
	 - sustaining long rule sequences in classifier systems
	 - modularizing classifier systems
	 - relations between evolutionary systems and neuromorphic systems
	 - search space assumptions of distributed search procedures

	There will be no proceedings at the workshop itself. However we are
	working with a publisher to publish a compilation volume of reviewed
	material based on the work presented at the workshop. Papers to appear
	in the compilation volume will be rigorously reviewed by a review
	committee yet to be appointed. Papers should be based on material
	presented at the workshop, but are not limited only to work presented
	there.

	A program committee consisting of Kenneth De Jong, David Goldberg,
	Gregory Rawlins, and Dirk Van Gucht will review submitted abstracts.
	Successful authors will be invited to attend the workshop. The number
	of attendees will be kept to at most 50.

	Abstracts should be no more than 15 pages long (excluding figures and
	tables), double-spaced, and set in 11pt or better.

	Deadline for submission is March 1st, 1990. Authors will be
	notified of the committee's decision by April 2nd, 1990.

	Please send four copies to:
	Gregory Rawlins
	Department of Computer Science
	101 Lindley Hall
	Indiana University
	Bloomington, Indiana 47405, USA

	rawlins@iuvax.cs.indiana.edu
	(812) 855-2136

	Thank you.


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

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

