
Genetic Algorithms Digest   Monday, March 18 1991   Volume 5 : Issue 5

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

Today's Topics:
	- CORRECTION ON DATES OF ICGA 91
	- Index for GA-List now available
	- Conference Announcement: AISB 91
	- Papers and TRs available
	- CFP: Evolution and Chaos in Cognitive Processing, IJCAI-91 Workshop
	- Graduate studies on GA, where?

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

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

4th Intl. Conference on Genetic Algorithms (v4n17)           Jul 13-16, 1991
AAAI 91, National Conference on AI, Anaheim, CA              Jul 14-19, 1991
IJCAI 91, International Joint Conference on AI, Sydney, AU   Aug 25-30, 1991
AISB 91, Leeds, UK (v5n5)                                    Apr 16-19, 1991

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

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

From: Alan C. Schultz (GA-List Moderator)
Date: Tuesday, March 12 1991
Subject: Correction

    The dates listed in the above calendar for ICGA 91 were
    wrong previous to this issue.  The conference dates had been
    changed a while ago to better align ICGA 91 with AAAI 91.
    ICGA 91 is from July 13-16.

    Alan

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

From: Alan C. Schultz (GA-List Moderator)
Date: Tuesday, March 12 1991
Subject: GA-list Index

    OK, the peer pressure has finally convinced me...

    I have put together an index of all previous GA-List
    Digests.  The index includes the issue number and
    the topics list from each issue.  If you would like
    a copy of the index, please request it from ga-list-request.
    FYI, there are currently 87 back issues of GA-List.

    Alan

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

From: B M Smith <bms@dcs.leeds.ac.uk>
Date: Fri, 22 Feb 91 14:59:18 GMT
Subject: Conference Announcement: AISB 91


                        *****************
                        *               *
                        *  A I S B 9 1  *
                        *               *
                        *****************
 
                     UNIVERSITY OF LEEDS, UK

                       16 - 19 APRIL 1991

                   TUTORIAL PROGRAMME 16 APRIL

                 TECHNICAL PROGRAMME 17-19 APRIL
                        with sessions on:
                  * Distributed Intelligent Agents
                  * Situatedness and Emergence in Autonomous Agents
                  * New Modes of Reasoning
                  * The Knowledge Level Perspective
                  * Theorem Proving
                  * Machine Learning

      Programmes and registration forms are now available from:
                  Barbara Smith
                  AISB91 Local Organizer
                  School of Computer Sudies
                  University of Leeds
                  Leeds LS2 9JT, UK

                  email: aisb91@ai.leeds.ac.uk

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

From: vose@cs.utk.edu
Date: Wed, 27 Feb 91 17:53:18 -0500
Subject: Papers and TRs available

    Recently a series of papers addressing GAs (and classifier systems)
    have been completed. Copies are available by surface mail.
    Please send full mailing address and titles of papers requested to
    gxl@msr.epm.ornl.gov (Gunar Liepins) or vose@cs.utk.edu (Michael Vose)

    1. Representational Issues in Genetic Algorithms 
    -- appeared in JETAI --  .
    Provides a constructive proof of fully deceptive 
    functions for n>2 and indicates that linear transformations 
    can transform these fully deceptive functions into fully
    easy functions.  Proves that other commonly considered 
    transformations are generally inadequate.

    2.  Punctuated Equilibria in Genetic Search.  
    Explicitly describes the (expected) dynamics of the GA 
    in terms of a fitness matrix and a mixing matrix.
    This paper formalizes GAs with an infinite population
    model incorporating selection, crossover, and mutation.
    Basic properties of the model are discussed and the
    qualitative results are applied to an explanation of
    punctuated equlibria.

    3.  Modeling Genetic Algorithms With Markov Chains.
    A simple genetic algorithm is modeled as a Markov chain.
    The method is complete in the sense that selection, mutation
    and crossover are both incorporated, and is exact in the sense
    that no special assumptions are made which restrict populations
    or population trajectories.  Asymptotics are also considered
    which show the relationship between real GAs (i.e. the Markov
    model) and the model given in 2.

    4.  Deceptiveness and Genetic Algorithm Dynamics.  
    Summarizes 1. and 2., reviews Bethke's deceptiveness 
    results, and introduces basis sets, a technique to generate 
    functions of arbitrary (but controlled) deceptiveness.  

    5.  Polynomials, Basis Sets, and Deceptiveness in Genetic Algorithms.
    Explores uses of and relationships between fitness values,
    polynomial forms, Walsh polynomials, and schemata utilities
    (basis sets).  Explores how these representations transform under 
    affine transformations of the search space. 

    6.  Characterizing Crossover in Genetic Algorithms.  
    Mathematically characterizes crossover in terms of interaction
    with schemata, translation, and projection.  Basic properties
    are established, and the extent of exploration which is driven
    by crossover is also determined.

    7.  apGA:  An Adaptive Parallel Genetic Algorithm.  
    This variant of the standard generational GA has solved all 
    (but DeJong's F4) test problems that have been presented to 
    it, including deceptive problems, interleaved deceptive 
    problems of mixed orders, and hierarchically deceptive 
    problems.  Unique to the algorithm is adaptive exponential 
    function deformation and limited temporal memory.  Adaptive 
    mutation is also used.  Much additional work remains 
    to resolve some of the seemingly arbitrary choices that 
    were made in the design of the algorithm.  Nonetheless, 
    apGA demonstrates that there is "more than one way to 
    skin the cat" 

    8.  Classifier System Learning of Boolean Concepts.  
    Matches population sizes and classifier generality to 
    the problem, uses crude speciation and queries to solve 
    the 20-multiplexor.  Sketches a proof that suggests that 
    classifer systems cannot be more efficient than Valiant's 
    "crossing out" algorithm for Boolean concept learning.  

    9. Generalizing The Notion Of Schemata In Genetic Algorithms.
    The concept of schema and the Schema Theorem are interpreted
    from a new perspective.  This allows GAs to be regarded as a
    constrained random walk, and offers a view which is amenable to
    generalization.

    10. Isomorphisms of Genetic Algorithms.
    The beginning premise is that the role of Holland schemata
    in directing genetic search should be granted both as
    empirical fact and as a natural consequence of the Schema
    Theorem.  The final conclusion is that schemata more general than
    Holland's can also be made to direct genetic search, and that
    a duality exists between problem representations and which
    schemata are relevant for their optimization.

    11.  Crossover Disruption.
    Results from the papers "Generalizing The Notion Of Schemata
    In Genetic Algorithms" and "Isomorphisms of Genetic Algorithms"
    are used to motivate a reexamination of schemata disruption.
    A ``building block hypothesis'' is formalized in which the relation of
    building blocks to utilities is not prominent.  What does matter
    is how the crossover operator interacts with schemata.  It is
    argued that this interaction determines what the building blocks are,
    and once determined, these building blocks are the appropriate
    objects on which to conduct schemata analysis.

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

From: dld@scs.carleton.ca (Dwight Deugo)
Date: Fri, 8 Mar 91 13:18:57 EST
Subject: CFP: Evolution and Chaos in Cognitive Processing, IJCAI-91 Workshop

                            IJCAI-91 Workshop

                          Call For Participation

                 Evolution and Chaos in Cognitive Processing 

    Goals

	Recently, there has been considerable interest and progress in the 
    study of systems of entities which, using only a simple set of local 
    rules, exhibit complex and robust global behavior. Much of this 
    activity has occurred in specialized research communities 
    investigating Artificial Life, Genetic Algorithms, Chaos Theory, and 
    Non-Linear Complex Systems. It is clear that many of the issues 
    being addressed by these groups have a strong intersection with the 
    goals and interests of the AI community - e.g. evolution and 
    behavior.  
	The goal of the Workshop is to provide a forum for researchers in 
    evolutionary processes and discrete chaos theory to investigate the 
    links between these two fields with respect to AI and cognitive 
    processing. This investigation is a first step towards answering the 
    following questions: Is an evolutionary model of cognition plausible? 
    How can correct, simple rules be determined which achieve a desired 
    global behavior? Is chaos theory a useful tool for the design and 
    analysis of evolutionary systems such as genetic algorithms and 
    classifier systems?

	Topics of interest for the Workshop include:

	    * Genetic algorithms	
	    * Classifier systems	
	    * Artificial life 
	    * Discrete chaos
	    * Discrete non-linear systems
	    * Cellular automata
	    * Emergent behaviors
	    * Non-classical evolutionary systems 
	    * Evolutionary epistemology
	    * The application of the above topics to cognitive processing

    Format

	All accepted papers will be presented in either a plenary or poster 
    session. In order to provide ample opportunity for discussion, most 
    presentations will be posters. Panel discussions will summarize the 
    findings of each area in the workshop and identify open problems 
    and future research directions. 

    Submissions

	Intended authors are invited to submit either a full paper (max. 
    15 pages), or a short paper (3-5 pages); all other participants should 
    submit a summary of previous relevant work with expected 
    contributions. All accepted papers will appear in the workshop 
    proceedings; the full papers will be considered for inclusion in a 
    planned book. Five copies should be submitted by May 13, 1991 to:

	Workshop on Evolution and Chaos in Cognitive Processing
	c/o Dwight Deugo
	School of Computer Science, 
	Carleton University, Ottawa, 
	Canada, K1S 5B6,  
	(613) 788-4333, 
	FAX (613) 788-4334
	e-mail: dwightdeugo@scs.carleton.ca

	Authors will be notified of the committee's decision by June 17, 
    1991. The final version will be due by July 15,1991.

    Organizing Committee

	Rob Black, Dwight Deugo, and Una-May O'Reilly (Carleton University)

    Program Committee

	Franz Oppacher and Nicola Santoro (Carleton University); Kenneth 
    De Jong (George Mason University); Christopher G. Langton (Los Alamos 
    National Laboratory)

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

From: pl160988@mtecv2.mty.itesm.mx (Ing. Ivan Ordonez)
Date: Thu, 7 Mar 91 21:13:16 CST
Subject: Graduate studies on GA

	 I am a graduate student about to finish a Master's degree on
    Computer Science. I intend to get a doctoral degree later, and I was
    wondering whether it is possible to major in Genetic Algorithms or a
    related field. If anybody has any information about places where such
    studies could be taken please let me know.

	    Ivan Ordonez-Reinoso. pl160988@mtecv2.mty.itesm.mx

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