Genetic Algorithms Digest    Tuesday, 30 August 1988    Volume 2 : Issue 19

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

Today's Topics:
	- GAs and regression
	- Booker's article in GAs and Sim. Annealing (2)
	- Next GA Conference (2)
	- Mailing lists
	- Update on Goldberg's GA Book (2)
	- Proceedings available from GA Conferences

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

Date: Wed, 10 Aug 88 09:30:43 -0400 (EDT)
From: David Greene <dg1v+@andrew.cmu.edu>
Subject: Re: GA-List v2n18 : "GA's and regression"

>Date: Wed, 27 Jul 88 11:12:57 EDT
>From: richardson <richards@UTKCS2.CS.UTK.EDU>
>Subject: GA's and regression
>
>Hello Ga World,
>
>We are planning on using a GA to calibrate a regression model.
>I have read De Jong and Smith, "Genetic Algorithms Applied to the Calibration
>of an Information Driven Model of U.S. Migration Patterns".
>
>Has anyone else applied GA's to regression?
>
>               Jon Richardson

You might check the paper Stephen and I wrote (Greene, D.P. and Smith, S.F. "A
Genetic System for Learning Models of Consumer Choice" 2nd  proceedings
ICGATA).

In it, a GA is compared with Quinlan's ID-3 and a linear Logit (regression)
model on a simulated data set.  The performance of both the GA and Logit was
quite good and highly comparable.  Perhaps of interest to you was that a
frequency measure of the features in the GA's rules correlated very highly with
the beta coefficients in the regression (many at p<.0001 significance).

I don't know if it's what you're looking for but you might find it
interesting.-David

dg1v@andrew.cmu.edu
Carnegie Mellon University

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

Date:         Thu, 11 Aug 88 14:27:05 CDT
From: "Robert Lingle, Jr." <PHLING@LSUVM.BITNET>
Subject:      Booker's article in GAs and Sim. Annealing

Lashon Booker's article in GAs & Simulated Annealing (ed. by L. Davis)
is missing TableII, while Table I has been printed twice. I would like
a copy of the real Table II. Maybe others would also.

Robert Lingle, Jr.
              Dept. of Physics and Astronomy / LSU
              Baton Rouge, LA   70803-4001

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

Date: Thu, 18 Aug 88 16:48:06 EDT
From: booker@nrl-aic.arpa
Subject: Re: Article in GA's & Simulated Annealing

   My paper in Dave Davis's book was inadvertently printed
without a copy of Table 2. Since that table summarized the
experimental results, I thought I'd make them available here.

Lashon

-------------------------------------------------------------------------------
|                                                                             |
|                        TABLE 2 - Performance Summary                        |
|              (Numbers in parentheses indicate that the  difer-              |
|              ence  in  performance  between this implementaion              |
|              and the previous one is not statistically signfi-              |
|              cant)                                                          |
|                                                                             |
+-----------------------------------------------------------------------------+
|       Crossover       |             Normalized Performance alues            |
|                       | -----------------------------------------------------
|    Implementation     |  On-Line|  Off-line|  Best-so-far| Weighted Off-line|
+---------------------------------+----------+-------------+------------------+
|                       |         |          |             |                  |
|One crossover point    |  3.3694 |   1.7260 |    0.1444   |      0.9014      |
|                       |         |          |             |                  |
|                       |         |          |             |                  |
|Two crossover points   |  3.5340 |   1.4119 |    0.0265   |      0.6739      |
|                       |         |          |             |                  |
|                       |         |          |             |                  |
|Constrained crossover  |  4.6410 |   1.2884 |   (0.0304)  |     (0.5888)     |
|                       |         |          |             |                  |
|                       |         |          |             |                  |
|Variable crossover rate|  4.1372 |   1.1540 |   (0.0238)  |      0.5137      |
-------------------------------------------------------------------------------

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

Date: Fri, 12 Aug 88 15:01:32 EDT
From: Rick_Riolo@um.cc.umich.edu
Subject: Next GA Conference

Some opinions:
1. I am neutral with respect to starting on Sunday.
   I can see where is would save some people money,
   and so might encourage some to come who might otherwise not.
2. I think the idea of having the paper presentations be more
   aimed at being discussions, rather than just rushed presentations,
   is a good one.  However, I think for this to be feasible people
   must have a chance to read the papers first.  To this end I
   think it would be worth having to submit papers one or two weeks
   earlier so that the Proceedings could be sent one or two weeks
   before the conference.
3. I think that publishing more papers than were presented
   at the Machine Learning conference was a reasonable solution
   to the problem of too many good papers to present in 3 days.
   If we can't hear it presented or discussed at the conference,
   at least we can know the work is being done by reading about it.
   And if we get the proceedings early, we could have a chance
   to talk to the authors at the conference.
  - r


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

Date: Thu, 18 Aug 88 17:00:38 PDT
From: Leonard <XT.A08@Forsythe.Stanford.EDU>
Subject: Next GA Conference

   I am interested in presenting a paper at the 3rd
International Conference on Genetic Algorithms. To
whom should I send an abstract?

   Thank you for your attention.

--------------------------------------------------------------------
L S Lutomski         Computer Systems Analyst
Palo Alto Office     American Institutes for Research
P. O Box #1113       1791 Arastradero Road
Palo Alto, CA 94302  415-493-3550
xt.a08 @ stanford

[ Dave Schaffer will serve as Program Chair for the next
GA Conference.  He will issue a Call for Papers very soon
that indicates deadlines for submitting papers.  The Call
be be included on GA-List, of course.  If you need more information
right now, contact:

	J. David Schaffer
        Philips Laboratories
        345 Scarborough Rd
        Briarcliff Manor, NY  10510

- JJG ]

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

Date:         Thu, 25 Aug 88 10:30:22 CST
From: Dave Goldberg <DGOLDBER@UA1VM.BITNET>
Subject:      CS and EE Department Mailing Lists in Machine Readable Form

     Does anyone know of a source for machine readable lists of Computer
Science Departments and Electrical Engineering Departments at colleges
and universities?  These would be helpful in publicizing the upcoming
GA Conference, and any information leading to their acquisition is
appreciated.  DEG

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

Date:         Thu, 25 Aug 88 14:38:31 CST
From: Dave Goldberg <DGOLDBER@UA1VM.BITNET>
Subject:      Printing Delay on GASOML

     Due to an unforeseen printing delay, my forthcoming book, GENETIC
ALGORITHMS IN SEARCH, OPTIMIZATION, AND MACHINE LEARNING, is still
forthcoming.  Addison-Wesley says that manufacturing should bind the book
on Friday, August 26 or Monday, August 29.  In any event, it should not
be long before copies are shipped to those who have ordered the book by
mail.  DEG

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

Date:         Mon, 29 Aug 88 15:08:43 CST
From: Dave Goldberg <DGOLDBER@UA1VM.BITNET>
Subject:      Update on GASOML Printing

     The final binding of GENETIC ALGORITHMS IN SEARCH OPTIMIZATION AND
MACHINE LEARNING has been delayed another two weeks.  Apparently, quality
of the cover printing was not up to par, and it must be redone.  AW says
that the book should be available on or around September 12.  I am sorry
for any inconvenience this may cause. DEG

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

Date: Tue, 30 Aug 88  9:15:06 EDT
From: John Grefenstette <gref@nrl-aic.arpa>
Subject: GA Conference Proceedings

The proceedings from both the 1985 and the 1987 GA Conferences
are now available from LEA.  Contact:

	Lawrence Erlbaum Associates, Publishers
	365 Broadway
	Hillsdale, NJ 07642
	USA
	(201) 666-4110

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

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