
Genetic Algorithms Digest   Thursday, September 17 1991   Volume 5 : Issue 32

 - 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
	- GA in Chess
	- RE: Long Genomes - an alternate explanation
	- CFP: FOGA-92 (Foundations of Genetic Algorithms)
	- Errata for Smith (1988) and Goldberg and Smith (1987).

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

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
 COGANN, Combinations of GAs and NNs, @ IJCNN-92 (v5n31)      Jun 6,     1992
 10th National Conference on AI, San Jose, (CFP 1/15)         Jul 12-17, 1992
 FOGA-92, Foundations of Genetic Algorithms, Colorado (v5n32) Jul 26-29, 1992
 ECAI 92, 10th European Conference on AI (v5n13)              Aug  3-7,  1992
 Parallel Problem Solving from Nature, Brussels, (v5n29)      Sep 28-30, 1992

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

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

From: jrv@sdimax2.mitre.org
Date: Fri, 27 Sep 91 18:46:54 EDT
Subject: Re:  order operators

   >>From: starkwea@CS.ColoState.EDU (tim starkweather)
   >>   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. 

   > This is a nice analysis, but it has a problem: it does not work for both
   > children. 

   True.  In my work only one child is kept, so that point had not
   occurred to me.

   > Whether only one child or both children should be kept is an interesting
   > question. Note that if masks are generated with a probability other than
   > .5, and only one child is kept, then P1 x P2 is not equivalent to P2 x P1.

   ...but even more basically, of course the two children of P1 x P2 are
   not equivalent.  If one child inherits one good building block from
   each parent, then its sibling necessarily inherits neither of those
   building blocks.  If children replace parents, obviously both children
   should be kept so that building blocks are only lost through
   disruption (and can in principle be regained).  However, if parents are
   retained (and evaluations dominate the GA overhead), I see no strong
   argument for evaluating the second child in preference to evaluating
   the child of another, independent, crossover.

   >This is because parents generally have a higher than average fitness, and
   >after P1 is chosen, the population P2 is chosen from has a lowered average
   >fitness, resulting in P1 having a higher average fitness than P2. Since the
   >mask is biased, the children will also have different average fitnesses.
   >This can be fixed by randomly selecting which child to keep, or by allowing
   >duplicates in the population and allowing P1 to equal P2 (or keeping both
   >children).

   ...or by using a symmetric selection procedure: choosing P1 and P2
   independently and rejecting the pair if they happen to be equal.

   On the other hand, suppose: 
	   P1 is expected to be substantially better than P2 (e.g. in my GAs,
	   one parent is chosen with a uniform distribution),

	   Both children are to be kept, and

	   Masks are generated with a probability greater than .5.

   One could arrange that inheritance will be mainly from the better
   parent by changing the way the two children are distinguished.  For the
   first child, let the masked points indicate the elements in P1 whose
   POSITION is inherited from P1.  For the second child, let the mask
   indicate the elements in P1 whose ORDER is inherited from P1.
   Alternately: the second child is generated by complementing the mask
   rather than exchanging the roles of the two parents.

   >A minor point: the term "crossover point" already has a clearly
   >defined meaning in the case of 1, 2, and n-point crossover, and depicts a
   >place between loci. It may be better to use another term, such as "mask,"
   >to indicate crossover operators that apply to loci instead of the breaks
   >between them.

   The use of "crossover point" for order- or position- based crossover
   operators made me uncomfortable too, though I don't think it necessary
   to introduce the term "mask operator" as Dr.  Syswerda may be
   suggesting.  I don't mind referring to any of these operators as
   crossover operators.  For 1, 2, and n-point crossover one needs to
   select "crossover points", whereas for order- or position- based
   crossover one needs to select a "mask".

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

From: gatkinso@NMSU.Edu
Date: Fri, 27 Sep 91 14:05:51 MDT
Subject: GA in Chess

   I am interested in the application of GAs to the formation of
   chess concepts by means of inductive example (I recently com-
   pleted a book-length manuscript with the working title *Chess
   and Machine Intuition*), and would like to establish contact
   with other enthusiasts.
   George Atkinson <gatkinso@nmsu.edu>

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

From: jrv@sdimax2.mitre.org
Date: Fri, 27 Sep 91 19:24:16 EDT
Subject: RE: Long Genomes - an alternate explanation

   There was a very interesting story in Analog a few months back wherein
   a researcher was demonstrating some really amazing technology
   (antigravity, etc.) and some government people were trying to hush it
   up - but it turned out he had deciphered the instructions from human
   DNA.  That is, the extra DNA is a message left by some extraterrestrial
   who passed through long ago - in a safe place he _knew_ we'd be looking
   eventually.

   I trust somebody on the human genome project has read the story.

			    - Jim Van Zandt (jrv@mbunix.mitre.org)

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

From: whitley@CS.ColoState.EDU (darrell  whitley)
Date: Fri, 27 Sep 91 09:21:40 MDT
Subject: CFP: FOGA-92 (Foundations of Genetic Algorithms)

%  The following is a latex version of the call for paper for FOGA-92

\documentstyle[foga]{article}

\title{FOUNDATIONS OF GENETIC ALGORITHMS 1992}

\author{\And
{\bf{}FOGA-92}\\
\AND }

\begin{document}

\maketitle

\vskip .5pc
\vskip .5pc

FOGA-92 will be held in the Colorado Rocky Mountains
in Vail, Colorado, July 26-29, 1992.
The purpose of the FOGA workshops is stimulate research on the
theoretical foundations of genetic algorithms.  

\vskip .5pc
{\bf CALL FOR PAPERS}
\vskip .5pc

A short version of FOGA papers will be due December 1, 1991.
This draft should not exceed 10 pages (using 12 point fonts)
and should describe specific results or work in progress.
Papers that exceed the 10 page limit will be rejected without
review.   Papers should focus on theoretical foundation
issues related to the computational behavior of genetic algorithms. 
Papers with an empirical or experimental orientation 
will be considered, but preference will be given to theoretical papers. 

Acceptance of the short version of papers will involve
oral or poster presentation at FOGA-92;  long versions of
accepted papers (upto 20 pages) will be due March 1, 1992 
and will be strongly considered for possible publication in a
book of select papers to be published by Morgan Kauffman,
similar to the first ``Foundations of Genetic Algorithms"
edited by Gregory Rawlins. 

\vskip .5pc
{\bf LIMITED PARTICIPATION}
\vskip .5pc

Space will be limited at FOGA-92 to approximately 60 participants.     
An effort will be made to invite all those who submit papers
to participate.
Individuals interested in attending this workshop who do not
plan to submit papers should send email to Darrell Whitley
(whitley@cs.colostate.edu)
for more information as soon as possible 
(and no later than December 1, 1991), or write to:
Denise Hallman, FOGA-92, Department of Computer Science,
Colorado State University, Fort Collins, CO 80523.

\end{document}

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

From: @ua1ix.ua.edu:rob@galab2.mh.ua.edu (Rob Smith)
Date: Tue, 01 Oct 91 16:09:06 CDT
Subject: Errata for Smith (1988) and Goldberg and Smith (1987).

   Hi,
    Recently, Dipankar Dasgupta at The University of Strathclyde pointed out
   errors in the problem description in my Master's thesis and related
   papers. Fortunately, the errors are slight and are only in the
   description, and do not affect any of the qualitative or quantitative
   results in these documents. However, I feel compelled to issue an errata
   to enable duplication of my experiments. The errata effect the following
   documents:

@techreport{Smith:88,
    author = "Smith, R. E.",
      year = "1988", 
     title = "An investigation of diploid genetic algorithms for adaptive search of nonstationary functions",
      titlenote = "Master's Thesis",
      type = "{TCGA} Report No.",
      number = "88001",
      institution= "The University of Alabama, The Clearinghouse for
Genetic Algorithms",
   address = "Tuscaloosa"}

@article{Goldberg:87g,
    author = "Goldberg, D. E. and Smith, R. E.",
      year = "1987",
     title = "Nonstationary function optimization using genetic
algorithms with dominance and diploidy",
   journal = "Proceedings of the the Second International Conference on
Genetic Algorithms",
     pages = "59--68"}

@article{Smith:87,
    author = "Smith, R. E.",
      year = "1987", 
     title = "Diploid genetic algorithms for search in time varying environments",
   journal = "Proceedings of the 25th Annual Southeast Regional Conference of the ACM", 
     pages = "175--178"}

   These documents indicate that a time-varying knapsack problem is used
   where

   "... the weight constraint is switched every 15 generations between
   .8 times the total weight of the available objects ... and .5 times
   the total weight"
	      (Smith, 1988, p. 40)

   Actually, the weight constraint was switched between 104 (about .86 times
   the total weight) and 60 (about .5 times the total weight).  It is likely
   that the error was a typo in an early version that propagated into the
   final documents.

   There is another typo in the table of problem parameters in the thesis
   (page 39).  Specifically, object 4 has value 2 rather than 1. This error
   does not appear in Goldberg and Smith (1987).

   Finally, the optimal strings listed in Goldberg and Smith (1987, page 66)
   contain a one-bit typo. The correct optima are shown in the thesis.  I
   believe I announced this error at ICGA2.

   To sum up, the following table (TeX format) contains a complete,
   corrected problem description that includes the correct optimal strings:

\begin{center}
\begin{tabular}{c|c|c|c|c}
\hline
	&		&		&	 $W = 60$ &	 $W = 104$\\
Number	&	Value	&	Weight	&	Optimal &	Optimal\\
$i$	&	$v_i$	&	$w_i$	&	$x_i$	&	$x_i$\\
\hline
1	&	2	&	12	&	0	&	0\\
2	&	3	&	5	&	1	&	1\\
3	&	9	&	20	&	0	&	1\\
4	&	2	&	1	&	1	&	1\\
5	&	4	&	5	&	1	&	1\\
6	&	4	&	3	&	1	&	1\\
7	&	2	&	10	&	0	&	0\\
8	&	7	&	6	&	1	&	1\\
9	&	8	&	8	&	1	&	1\\
10	&	10	&	7	&	1	&	1\\
11	&	3	&	4	&	1	&	1\\
12	&	6	&	12	&	1	&	1\\
13	&	5	&	3	&	1	&	1\\
14	&	5	&	3	&	1	&	1\\
15	&	7	&	20	&	0	&	1\\
16	&	8	&	1	&	1	&	1\\
17	&	6	&	2	&	1	&	1\\
\hline
Total:	&	91	&	121	&	13	&	15\\
\hline
\multicolumn{3}{c|}{\ }	&	 $\sum_{i=1}^{17} x_i v_i = 71$ &	 $\sum_{i=1}^{17} x_i v_i = 87$\\
\multicolumn{3}{c|}{\ }	&	 $\sum_{i=1}^{17} x_i w_i = 59$	&	 $\sum_{i=1}^{17} x_i w_i = 99$\\
\cline{4-5}
\end{tabular}
\end{center}

   As I have said, none of these errors affect the results in these
   documents. All experiments used the problem described above.

    Robert Elliott Smith
    Department of Engineering of Mechanics
    The University of Alabama
    P. O. Box 870278
    Tuscaloosa, Alabama 35487
<<email>> @ua1ix.ua.edu:rob@galab2.mh.ua.edu 
<<phone>> (205) 348-1618
<<fax>> (205) 348-8573    

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

