Genetic Algorithms Digest    Friday, 26 February 1988    Volume 2 : Issue 4

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

Today's Topics:
	- Question about genetic algorithms
	- Software
	- Discussion from AIList

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

Date: Tue, 19 Jan 88 14:40:48 EST
From: Lashon Booker <booker@NRL-AIC.ARPA>
Subject: Re: Question about genetic algorithms
[Concerning Tom Dietterich's question in v2n3 -- JJG]

Someone has probably already told you this by now, but in
case they haven't, John Holland's original book (1975)
contains a brief analysis of a problem similar to the one
you describe. Assuming a *FIXED* sized population subject
to repeated crossover with uniform random pairing (the two
offspring replace their parents), the distribution of schema
reaches a stochastic steady state. In this equilibrium
distribution the probability of occurence of *every* schema is
the product of the proportions of its defining alleles. In
particular, the probability of occurence of any individual is the
product of the proportions of all its alleles. The expected interval
between occurences of an individual is given by the reciprocal of
this probability.
Holland's analysis starts at the bottom of page 99.

Lashon

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

Date: 27 Jan 88 12:03 -0700
From: Kristinn Kristinsson <kristinnk%ppc.ubc.cdn@ubc.CSNET>
Subject: Software

I am trying to use Genetic Algorithm in adaptive control.

Is it possible to get a list of available GA software and where to
get it from?

Thanks in advance,

Kristinn Kristinsson
Pulp and Paper Centre
University of British Columbia
2385 East Mall
Vancouver, B.C. V6T 1K2
Canada


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

Date: Thu, 25 Feb 88 13:11:32 EST
From: John Grefenstette <gref@NRL-AIC.ARPA>
Subject: Discussion of GA's on AIList

[The following discussion recently appeared on AIList.
It is included here for those of you who don't subscribe
to that list, or who may have missed it. -- JJG]

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

Date: 25 Jan 88 06:39:33 GMT
From: g451252772ea@deneb.ucdavis.edu
Subject: Re: Cognitive System using Genetic Algorithms.

About a month or so ago I complained of the engineering focus of disserta-
tions done by Holland's students.  I got a very nice reply from a former
student, Lashon Booker, who cited a number of more abstract projects
(including his own).  All these theses are available through U. Microfilms
(which happens to be based in Michigan), at about $25 each.

Lashon is still quite active; he's at booker@nrl-aic.ARPA.  There is a BBS
for genetic algorithms; to subscribe, send mail to GA-List-Request@nrl-aic.ARPA.
(I did some time ago but have no reply yet... hmmm)

	[ GA-List had a wrong address.  Sorry -- JJG]

And a standard set of C subroutines for classifier systems is available for
media cost from Rick Riolo at U.Mich.  Contact him at
Rick_Riolo@ub.cc.umich.edu for details - I got mine on a 1.2 meg AT disc (just
fits).  Other formats available (Sun, Mac, ... ).  This is ver 0.98, so it's
not totally stable yet.  I'm slowly getting acquainted with it all...

Oh yes: the books INDUCTION, 1986, by Holland et al; GENETIC ALGORITHMS AND
SIMULATED ANNEALING, 1987, L. Davis; and GENETIC ALGORITHMS AND THEIR
APPLICATION, Proceed. 2nd Intl. Conf. Gen. Alg. (L. Erlbaum Assoc, Pub),
are all of interest.

I, for one, would be curious what else you learn, although my interests are
more in the theoretical arena (population genetics, et al).


Ron Goldthwaite / UC Davis, Psychology and Animal Behavior
'Economics is a branch of ethics, pretending to be a science;
 ethology is a science, pretending relevance to ethics.'

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

Date: 2 Feb 88 22:02:00 GMT
From: goldfain@osiris.cso.uiuc.edu
Subject: Re: Cognitive System using Genetic Algo


Would someone do me a favor and post or email a short definition of the
term "Genetic Learning Algorithm" or "Genetic Algorithm" ?

Thanks.    - Mark Goldfain  arpa:  goldfain@osiris.cso.uiuc.edu

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

Date: Fri, 5 Feb 88 10:08:02 PST
From: rik@sdcsvax.ucsd.edu (Rik Belew)
Subject: A short definition of Genetic Algorithms

Mark Goldfain asks:
        Would someone do me a favor and post or email a short definition of the
        term "Genetic Learning Algorithm" or "Genetic Algorithm" ?

I feel like Genetic Algorithms has two, not quite distinct meanings
these days. First, there is a particular (class of) algorithms developed
by John Holland and his students. This GA(1) has at its most distinctive
feature the "cross-over" operator, which Holland has gone to some
effort to characterize analytically. Then there is a broader class GA(2)
of genetic algorithms (sometimes also  called "simulated evolution") that
bear some loose resemblence to population genetics. These date back
to at least Fogel, Owen and Walsh (1966). Generally, these
algorithms make use of only a "mutation" operator.
        The complication comes with work like Ackley's thesis (CMU, 1987)
which refers to Holland's GA(1), but which is most accurately
described as a GA(2).

Richard K. Belew

rik@cs.ucsd.edu

Computer Science & Engr. Dept.  (C-014)
Univ. Calif - San Diego
San Diego, CA 92093

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

Date: 5 Feb 88 18:22:21 GMT
From: g451252772ea@deneb.ucdavis.edu  (0040;0000003980;0;327;142;)
Subject: Re: Cognitive System using Genetic Algo


I offer definitions by (1) aspersion (2) my broad characterization (3) one
of J Holland's shortest canonical characterizations and (4) application.

(1)  GA are anything J Holland and/or his students say they are.  (But this
_is_ an aspersion on a rich, subtle and creative synthesis of formal systems
and evolutionary dynamics.)

(2) Broadly, GA are an optimization method for complex (multi-peaked, multi-
dimensional, ill-defined) fitness functions.  They reliably avoid local
max/min, and the search time is much less than random search would require.
Production rules are employed, but only as mappings from bit-strings (with
wild-cards) to other bit strings, or to system outputs.  System inputs are
represented as bitstrings.  The rules are used stochastically, and in
parallel (at least conceptually; I understand several folk are doing
implementations, too).

A pretty good context paper for perspective (tho weak on the definition of
GA!) is the Nature review 'New optimization methods from physics and
biology' (9/17/87, pp.215-19).  The author discusses neural nets,
simulated annealing, and one example of GA, all applied to the TSP, but
comments that "... a thorough comparason ... _would be_ very interesting"
(my emphasis).

(3)  J. Holland, "Genetic algorithms and adaptation", pp. 317-33 in
ADAPTIVE CONTROL OF ILL-DEFINED SYSTEMS, 1984, Ed. O. Selfridge, E. Rissland,
M. A. Arbib.  Page 319 has:
"In brief, and very roughly, a genetic algorithm can be looked
upon as a sampling procedure that draws samples from the set C; each
sample drawn has a value, the fitness of the corresponding genotype.
From this point of view the population of individuals at any time t,
call it B(t), is a _set_ of samples drawn from C.  The genetic algo-
rithm observes the fitnesses of the individuals in B(t) and uses
this information to generate and test a new set of individuals,
B(t+1).  As we will soon see in detail, the genetic algorithm uses
the familiar "reproduction according to fitness" in combination with
crossing over (and other genetic operators) to generate the new
individuals.  This process progressively biases the sampling pro-
cedure toward the use of _combinations_ of alleles associated with
above-average fitness.  Surprisingly, in a population of size M, the
algorithm effectively exploits some multiple of M^3 combinations in
exploring C.  (We shall soon see how this happens.)  For populations
of more than a few individuals this number, M^3, is vastly greater
than the total number of alleles in the population.  The correspond-
ing speedup in the rate of searching C, a property called _implicit
parallelism_, makes possible very high rates of adaptation.  Moreover,
because a genetic algorithm uses a distributed database (the popu-
lation) to generate new samples, it is all but immune to some of the
difficulties -- false peaks, discontinuities, high-dimensionality,
etc. -- that commonly attend complex problems."

Well, _I_ shall soon close here, but first the few examples of applications
that I know of (the situation reminds me of the joke about the two rubes
visiting New York for the first time, getting off the bus with all of
$2.50.  What to do?  One takes the money, disappears into a drugstore
and reappears having bought a box of Tampax.  Quoth he, "With tampax,
you can do _anything_!)  Anyway:

o       As noted, the TSP is a canonical candidate.
o       A student of Holland has implemented a control algorithm for
a gas pipe-line center, which monitors and adaptively controls flow
rates based on cyclic usages and arbitrary, even ephemeral, constraints.
o       Of course, some students have done some real (biological) population
genetics studies, which I note are a tad more plausible than the usual
haploid, deterministic equations.
o       Byte mag. has run a few articles, e.g. 'Predicting International
Events' and 'A bit-mapped Classifier' (both 10/86).
o       Artificial animals are being modelled in artificial worlds.  (When
will the Vivarium let some their animated blimps ("fish") be so programmed?)

Finally, I noted above that the production rules take system inputs as
bit-strings.  This representation allows for induction, and opens up a
large realm of cognitive science issues, addressed by Holland et al in
their newish book, INDUCTION.

Hope this helps.  I really would like to hear about other application
areas; pragmatic issues are still unclear in my mind also, but as apparent,
the GA model has intrinsic appeal.




Ron Goldthwaite / UC Davis, Psychology and Animal Behavior
'Economics is a branch of ethics, pretending to be a science;
 ethology is a science, pretending relevance to ethics.'

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

Date: 3 Feb 88 18:13:15 GMT
From: umich!dwt@umix.cc.umich.edu  (David West)
Subject: Re: Classifier System Testbed

In article <241@wright.EDU> joh@wright.EDU (Jae Chan Oh) writes:
>Does anyone know where Rick Riolo (a former grad. student at Univ. of
>Mich.) is located at present, or how can I reach him by email...

You should be able to reach him at  Rick_Riolo@ub.cc.umich.edu
(case not significant).
-David.

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

Date: 13 Feb 88 22:13:47 GMT
From: jason@locus.ucla.edu
Subject: becoming literate with genetic algorithms

John Holland was here recently giving talks on genetic algorithms.  I found the
concept rather intriguing.  After hearing his lectures, I realized I needed to
do some introductory reading on the subject to fully appreciate its potential.

I am particularly interested in getting some references in the following areas:

(1) introductory theory behind GA
(2) its application to rule-based learning systems
(3) its relation to and implementation as neural nets

Thanks,

Jason Rosenberg                      Mira Hershey Hall
                                     801 Hilgard Avenue
jason@cs.ucla.edu                    Los Angeles, CA  90024
{ihnp4,ucbvax}!ucla-cs!jason         (213) 209-1806

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

Date: 11 Feb 88 19:34:52 GMT
From: umich!dwt@umix.cc.umich.edu  (David West)
Subject: Re: Cognitive System using Genetic Algorithms

In article <1062@ucdavis.ucdavis.edu> g451252772ea@deneb.ucdavis.edu.UUCP
 (PUT YOUR NAME HERE) writes:
>The author discusses neural nets,
>simulated annealing, and one example of GA, all applied to the TSP, but
>comments that "... a thorough comparason ... _would be_ very interesting"
  [...]
>As noted, the TSP is a canonical candidate.

I believe the TSP is popular because it is easy and compact to program.
The performance of a general method such as GAs can be strongly influenced
by the problem representation, and it turns out that the most straightforward
representations for genetic operations are particularly badly matched to the
most straightforward representations for TSPs.  This makes the TSP a rather
unfortunate choice of introductory example for people who are unfamiliar
with GAs.

>Finally, I noted above that the production rules take system inputs as
>bit-strings.  This representation allows for induction,...

It is *one* way of getting a form of induction, and has the property that
only very simple operations on the internal representation are used; the
extent to which this is useful depends, again, on the joint appropriateness
of the representations of the genetic operators and the world.
An "appropriate" representation has the property that the expected
fitness of the result of (say) a crossover is not severely worse than
that of its parents. This is something that must be ensured by the experimenter
if (as is most common) the representational mapping itself is not subject to
genetic selection.

 -David West

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

Date: 15 Feb 88 18:22:11 GMT
From: g451252772ea@deneb.ucdavis.edu  (0040;0000001899;0;327;142;)
Subject: Re: becoming literate with genetic algorithms

The references, generally at good libraries, that I know of for GAs:

Introductory:
 Holland, J., et al.  INDUCTION.  1986, MIT Press.  The book is a coherent
whole, not a collection of separately authored papers - and reads very well
by any standard.  Most of it discusses human induction, but the main model
introduced early on is Holland's.  And the human material is fascinating in
its own right, only partly because of the lucid presentation.  The description
of Holland's GA is complete, and an alternative system, PI, is also presented.
This is a more familiar symbol-based production system, in LISP.

 Holland, J. "Genetic Algorithms and Adaptation", in O. Selfridge, et al,
ADAPTIVE CONTROL OF ILL-DEFINED SYSTEMS.  1984, Plenum Press, NY.  This is
a discrete chapter, in which an overview of GA is provided. Almost every main
theme is touched on.

 Davis, L.  GENETIC ALGORITHMS AND SIMULATED ANNEALING.  1987, Morgan Kauffman
Pub, Los Altos, CA.  A collection of research papers by Holland's colleagues,
mostly (his INDUCTION chapters are reproduced here also).  A good variety of
current work, and again very lucid as technical/research writing goes (by
contrast, the Neural net literature is hopeless).  Topics include a study of
the TSP; parallel implementation of the CFS-C simulation library for GA on
the Connection Machine (nice!); Axlerod's study of GA in round-robins of the
iterated Prisoner's dilemma; a somewhat vague but very suggestive study on
designing a mapping from 'an East Asian language' onto a usable keyboard,
using a GA; some formal tests of 'hard' problems for GAs; and another
suggestive paper (for me) on producing long action sequences with GA by
means of 'hierarchical credit allocation' (this problem has parallels in
the animal-behavior literature I'm familiar with).

 Holland, J.  ADAPTION IN NATURAL AND ARTIFICIAL SYSTEMS.  1975, U. Michigan
Press.  The definitive foundation, marred only by a generous use of formal
notation (not insensibly, but offputting nonetheless).  The main conceptual
addition since this has been the interpretive change in INDUCTION, I think.

The GA community has held two conferences, last summer and in '86.  The
proceedings are available from Lawrence Erlbaum Assoc., 365 Broadway,
Hillsdale, NJ 07642.  My copy is on order ("Proc. Second International
Conf. on GA and their applications", held at Cambridge, MA, July 28-81, 1987).

And the various dissertations Holland has supervised are worth perusing via
U.Microfilm copies at $25 each.

For relating GA to NNets, I'll hazard to volunteer Richard Belew's name.  He
responded to an earlier posting I made and stated an interest in what
commonalities there might be.  He teaches at UCSD: rik@sdcsvax.ucsd.edu.

Oh yes: as the _very best_ intro article to GA, I recommend the final issue
of Science 86, for July, I think.  Too bad that mag died.

Hopefully helpfully (let me know what else you find- I've been teaching this
material to budding animal behaviorists!) -



Ron Goldthwaite / UC Davis, Psychology and Animal Behavior
'Economics is a branch of ethics, pretending to be a science;
 ethology is a science, pretending relevance to ethics.'

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

Date: 14 Feb 88 15:31:54 GMT
From: trwrb!aero!venera.isi.edu!smoliar@ucbvax.Berkeley.EDU  (Stephen
      Smoliar)
Subject: Re: becoming literate with genetic algorithms

In article <9430@shemp.CS.UCLA.EDU> jason@CS.UCLA.EDU () writes:
>John Holland was here recently giving talks on genetic algorithms.  I found
>the
>concept rather intriguing.  After hearing his lectures, I realized I needed to
>do some introductory reading on the subject to fully appreciate its potential.
>
The best source would be the book entitled INDUCTION, which Holland wrote with
Holyoak, Nisbett, and Thagard.  Most of the material from the talk is in
Section 4.1 (I think).  The preceding material leading up to the major
argument is very well written, as is the subsequent discussion.

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



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


