Copy Link
Add to Bookmark
Report
Neuron Digest Volume 13 Number 33
Neuron Digest Tuesday, 14 Jun 1994 Volume 13 : Issue 33
Today's Topics:
Adaptive Simulated Annealing (ASA) version 3.13
Hard System for NN
solution to mind-body problem
WWW Neural Network Page at OFAI
Damaged NNs
Stats vs ANNs : A Competition
URGENT QUESTION
Research Opportunities Parallelizing ASA and PATHINT
Cognitive Science position at Exeter
Postdoctoral Position: Applying Machine Learning to Ecosystem Modeling
Research position available
PostDoc position
Fall School on Connectionism and Neural Nets HeKoNN 94 (in German)
Send submissions, questions, address maintenance, and requests for old
issues to "neuron-request@psych.upenn.edu". The ftp archives are
available from psych.upenn.edu (130.91.68.31) in pub/Neuron-Digest or by
sending a message to "archive-server@psych.upenn.edu".
----------------------------------------------------------------------
Subject: Adaptive Simulated Annealing (ASA) version 3.13
From: Lester Ingber <ingber@alumni.caltech.edu>
Date: Wed, 01 Jun 1994 09:18:50 -0700
Adaptive Simulated Annealing (ASA) version 3.13
The latest Adaptive Simulated Annealing (ASA) code and some related
(p)reprints can be retrieved via anonymous ftp from
ftp.alumni.caltech.edu [131.215.139.234] in the /pub/ingber directory.
A new OPTION now collects data during the global optimization process
to importance-sample the user's variables.
Interactively [brackets signify machine prompts]:
[your_machine%] ftp ftp.alumni.caltech.edu
[Name (...):] anonymous
[Password:] your_e-mail_address
[ftp>] cd pub/ingber
[ftp>] binary
[ftp>] ls
[ftp>] get file_of_interest
[ftp>] quit
The 00index file contains an index of the other files and information
on getting gzip and unshar for DOS, MAC, UNIX, and VMS systems.
The latest version of ASA, ASA-x.y (x and y are version numbers), can
be obtained in several formats. ASA-shar.Z is a compress'd shar'd file
of the current code. For the convenience of users who do not have any
uncompress/gunzip utility, there is a file ASA-shar which is an
uncompress'd copy of ASA-shar.Z; if you do not have sh or shar, you
still can delete the first-column X's and separate the files at the
END_OF_FILE locations. There are tar'd versions in compress and gzip
format, ASA.tar.Z and ASA.tar.gz, respectively. There also is a
current zip'd version, ASA.zip, in which all files have been processed
through unix2dos. Directory /pub/ingber/0lower.dir contains links to
these files for some PC users who may have difficulty with upper case.
If you do not have ftp access, get information on the FTPmail service
by: mail ftpmail@decwrl.dec.com, and send only the word "help" in the
body of the message.
If any of the above are not possible, and if your mailer can handle
large files (please test this first), the code or papers you require
can be sent as uuencoded compressed files via electronic mail. If you
have gzip, resulting in smaller files, please state this.
Sorry, I cannot assume the task of mailing out hardcopies of code or
papers. My volunteer time assisting people with their queries on my
codes and papers must be limited to electronic mail correspondence.
To get on or off the ASA_list e-mailings, just send an e-mail to
asa-request@alumni.caltech.edu with your request. Update notices are
sent to the ASA_list about every month or two, more frequently if
warranted, e.g., in cases of important bug fixes; these notices are the
only e-mail sent to the ASA_list.
Lester
|| Prof. Lester Ingber ||
|| Lester Ingber Research ||
|| P.O. Box 857 E-Mail: ingber@alumni.caltech.edu ||
|| McLean, VA 22101 Archive: ftp.alumni.caltech.edu:/pub/ingber ||
------------------------------
Subject: Hard System for NN
From: lorenzi@diogene.univ-lyon2.fr
Date: Fri, 03 Jun 1994 01:01:27 -0200
[[ Editor's Note: I'm not quite sure what this person is asking. Perhaps
there is eitehr a) a translation problem or b) a meaning of "hard" of
which I'm not aware? Perhaps some kind soul could help? -PM ]]
I'm trying to find any hard system on PC which could
simulate Neural Networks.
It's realy inportant for me, thanks
_______________________________________________________________________
I DURAND E. I I
Iat lorenzi@diogene.univ-lyon2.fr I I
I FRANCE I I
I_________________________________I____________________________________I
------------------------------
Subject: solution to mind-body problem
From: duch@phys.uni.torun.pl (Wlodzislaw Duch)
Date: Tue, 07 Jun 1994 13:31:37 +0200
[[ Editor's Note: At last! -PM ]]
I have found an interesting solution to the mind-body problem and would be
grateful for comments on the paper that will be submitted to PSYCOLOQUY
soon. Since the paper has about 46 kB I just give a reference to it.
Our anonymous ftp archive is on
class1.phys.uni.torun.pl
Login as ftp, give your id@node.name as password
cd pub/publications/kmk
Files are in postscript and are compressed (*.Z type) but the file
m-blong.tex is also stored as an ASCII file (Latex format).
Our site is accessible also via WWW (Mosiac) through Polish home page - look
for Torun at the list of WWW servers, not on the map (the map connects with
a gopher server), or give our machine address as URL.
Wlodzislaw Duch
duch@phys.uni.torun.pl
Department of Computer Methods
Nicholas Copernicus University
Grudziadzka 5, 87-100 Torun, Poland
Tel/Fax ++48-56-21543
------------------------------
Subject: WWW Neural Network Page at OFAI
From: Georg Dorffner <georg@ai.univie.ac.at>
Date: Wed, 08 Jun 1994 13:07:06 +0200
[[ Editor's Note: Please check with your local site experts on how to
access Mosiac, WWW, or other sources. There is much software for
Internet searching available, but it's beyond my scope to explain how to
install and use it. -PM ]]
=========================================================
Announcing a New WWW Page on Neural Networks
at the
Austrian Research Institute for Artificial Intelligence
=========================================================
As part of the Worl-Wide-Web (WWW) server of the Department of
Medical Cybernetics and Artificial Intelligence of the
University of Vienna and the Austrian Research Institute for
Artificial Intelligence (OFAI)
URL: http://www.ai.univie.ac.at
a home page specifically dedicated to our research and services
in neural networks has been established:
URL: http://www.ai.univie.ac.at/oefai/nn/nngroup.html
It gives a description of the research currently being
undertaken at the Neural Network Group of the OFAI, which
consists of the following four domains:
- practical applications of neural networks
- theoretical research on neural networks
- cognitive modeling with neural networks
- neural network simulation tools
A complete list of publications is given, many of which can be
directly retrieved as postscript files.
Among the local services provided you are welcome to use our
========================================
bibliographical search utility BIBLIO,
========================================
which permits you to search among 3500 books and papers in the
field of neural networks. Search key can be an author's name
and/or a string contained in the title. The basis for the search
is an on-line data base containing books, reports, journal and
conference references, such as IEEE and INNS neural network
conferences. This data base is constantly being extended.
Finally, links to other neural network WWW pages, as well as
data and report repositories are also given.
Enjoy!
P.S: If you have questions or difficulties, mail to
georg@ai.univie.ac.at
------------------------------
Subject: Damaged NNs
From: kort@olymp.informatik.uni-bonn.de
Date: Mon, 13 Jun 1994 04:04:09 -0700
Hello ! I'd posted a request concerning damaged NNs some month ago and
received some answers, as well as questions to publish the results.
So here they are
Lesioning an attractor network : investigations of acquiered dyslexia,
G.Hinton, T. Shallice in Psychological Review 1991, Vol.98, No1, p.74-95,
(appeared also Scientific American 1993).
4-2-4 Encoder Death, S.L. Thaler in 1993 World Congress on NN, Portland,
Oregon, July 11-15, p. 180-183.
Alexander Kort.
------------------------------
Subject: Stats vs ANNs : A Competition
From: Jim Kay <sassk@macaulay-land-use.scot-agric-res-inst.ac.uk>
Date: Mon, 13 Jun 1994 11:48:58 +0000
Statistics vs. Neural Networks
A Competition
Can artificial neural networks outperform statistical methods
in a fair comparison ?
Finding applications where they can is one of the main objectives of a
two-day workshop to be held on April 19-20, 1995 in Edinburgh, Scotland.
We invite entries to this competition which should reach Jim Kay at the address
given below by November the first. The decisions reached will be communicated
to applicants by the 15th of January, 1995.
The best four entries will be selected and one applicant per entry will be
invited to attend the workshop and make an oral presentation of their results;
costs of accommodation and travel (within the UK) will be provided subject to
certain upper bounds.
The other general objectives of the workshop are:
to discuss problems of statistical interest within ANN research;
to discuss statistical concepts and tools that expand the technology
of ANN research;
to enhance collaborative research involving experts from one or more of
the two communities.
We look forward to receiving your applications which should include
a contact name and address and be no more than 10 typed A4 pages.
Jim Kay and Mike Titterington
SASS Environmental Modelling Unit
Macaulay Land Use Research Institute
Craigiebuckler
Aberdeen AB9 2 QJ
Scotland, UK
e-mail : j.kay@uk.ac.sari.mluri (within the UK)
j.kay@mluri.sari.ac.uk (internet address)
Tel. : +224 - 318611 (ext. 2269)
Fax : +224 - 208065
------------------------------
Subject: URGENT QUESTION
From: mwitten@chpc.utexas.edu
Date: Mon, 13 Jun 1994 12:18:35 +0600
Question:
I am interested in any help from anyone out there on the
following subject.
We have become an information gathering
society. One of the areas of interest is the gathering of
large databases of information. I was wondering if
anyone out there might know of databases of biomedical
information and how to access them. For example,
GenBank, census data, databases of medical images,
databases of population statistics, environmental
toxicity data, dental data, etc. If you know of such data,
would you please send me the following information:
Name of Database:
Is the database public or private:
How to contact database or database owner.
I will summarize the responses for the list. This is a
rather urgent request so, rapid answers would be
appreciated. Do not worry about duplicating others.
Feel free to cross post this note to any other lists that
might be appropriate.
Thanks,
Matthew Witten, Ph.D.
Head, Department of Applications Research and Development
Associate Director
UT System Center For High Performance Computing
Balcones Research Center, 1.154 CMS
10100 Burnet Road, Austin, TX 78758-4497 USA
Phone: (512) 471-2472 FAX: (512) 471-2445
E-MAIL MWITTEN@CHPC.UTEXAS.EDU
------------------------------
Subject: Research Opportunities Parallelizing ASA and PATHINT
From: Lester Ingber <ingber@alumni.caltech.edu>
Date: Sun, 05 Jun 1994 10:15:26 -0700
Research Opportunities Parallelizing ASA and PATHINT
I am looking for one to several people with experience parallelizing C
code, e.g., on Crays, to work on parallelizing two specific
algorithms: (a) Adaptive Simulated Annealing (ASA), and (b) an
algorithm to calculate the time-development of multi-variable nonlinear
Fokker-Planck-type systems, using a powerful non-Monte Carlo path
integral algorithm (PATHINT). Some code and papers dealing with these
algorithms can be obtained from ftp.alumni.caltech.edu
[131.215.139.234] in the /pub/ingber directory.
I am PI of an award of Cray time on an NSF Supercomputer, and have
ported these codes successfully onto a C90. However, I am short of
time to further optimize these codes, which is an essential requirement
before doing production runs on C90 and T3D Crays. If necessary, I
will do this work myself, but I would rather share the work,
experience, and research with other interested people who also can
expedite these projects. All code will remain under my copyright under
the GNU General Public License (GPL), i.e., the least restrictive
Library GPL.
There are several immediate projects that are just waiting for detailed
calculations, using codes which already run on SPARCstations, but need
the power of supercomputers for production runs. All results of these
studies will be published in peer-reviewed scientific journals, and
only active participants on these projects will be co-authors on these
papers.
Examples of these projects include:
neuroscience
EEG correlates of behavioral states
short-term memory modeling
realistic chaos + noise modeling
financial applications
2- and 3-state term-structure security calculations
testing of trading rules
nonlinear modeling
persistence of chaos in the presence of moderate noise
If you are interested, please send me a short description of projects
you have worked on, and how many hours/week you are prepared to commit
to these projects for at least a period of 6-12 months.
Lester
|| Prof. Lester Ingber ||
|| Lester Ingber Research ||
|| P.O. Box 857 E-Mail: ingber@alumni.caltech.edu ||
|| McLean, VA 22101 Archive: ftp.alumni.caltech.edu:/pub/ingber ||
------------------------------
Subject: Cognitive Science position at Exeter
From: D.C.Mitchell@exeter.ac.uk
Date: Wed, 08 Jun 1994 12:11:44 +0100
Connectionist (especially language & connectionist) applications
are invited for the attached job. Apologies if this has been
posted before (I'm not on the list!)
Don Mitchell
----------------------------------------------------------
Department of Computer Science
Lectureship in Computer Science (Cognitive Science)
The Department of Computer Science, recently rated excellent by HEFCE
teaching quality assessment, requires:
LECTURER (COGNITIVE SCIENCE) (ref. 3913e)
from September or as soon as possible. Candidates should have
research and teaching interests in computational aspects of cognitive
science (e.g. artificial intelligence, neurocomputing, visualization,
computational linguistics, computational modelling of mind, cognitive
architecture). Candidates should have an appropriate PhD and be able to
contribute to existing research and teaching areas in the
Department and to the BSc Single Honours interdisciplinary degree in
Cognitive Science (with the Department of Psychology). Salary up to
15,796 pounds p.a. on Lecturer A scale 13,601-18,855 pounds p.a.
(under review).
Information from Personnel, University of Exeter, Exeter EX4 4QJ;
(0392) 263100, or email Personnel@exeter.ac.uk, quoting the above
reference number. Closing date 6 July 1994.
Informal enquiries should be directed towards: Ajit Narayanan,
Department of Computer Science, (0392) 264064, or email:
ajit@dcs.exeter.ac.uk
(Equal Opportunities Employer)
------------------------------
Subject: Postdoctoral Position: Applying Machine Learning to Ecosystem Modeling
From: Tom Dietterich <tgd@chert.CS.ORST.EDU>
Date: Wed, 08 Jun 1994 12:48:02 -0700
Postdoctoral Position: Applying Machine Learning to Ecosystem Modeling
Complex ecosystem models are calibrated by manually fitting them to
available data sets. This is time-consuming, and it can result in
overfitting of the models to the data. We are applying machine
learning methods to automate this calibration and thereby improve the
reliability and statistical validity of the resulting models. Our
ecosystem model--MAPSS--predicts amounts and types of vegetation that
will grow under global warming climate scenarios. An important goal
of global change research is to incorporate such vegetation models
into existing ocean-atmosphere physical models.
Under NSF funding, we are seeking a Post-Doc to assume a major role in
carrying out this research. Components of the research involve (a)
representing ecosystem models declaratively, (b) implementing
gradient and non-gradient search techniques for parameter fitting,
(c) implementing parallel algorithms for running and fitting the
ecosystem model, and (d) conducting basic research on issues of
combining prior knowledge with data to learn effectively. The ideal
candidate will have a PhD in computer science or a closely related
discipline with experience in neural networks, simulated annealing
(and similar search procedures), knowledge representation, and
parallel computing. The candidate must know or be eager to learn some
basic plant physiology and soil hydrology. Computational resources
for this project include a 16-processor 1Gflop Meiko multicomputer and
a 128-processor CNAPS neurocomputer.
Applicants should send a CV, summary of research accomplishments,
sample papers, and 3 letters of reference to
Thomas G. Dietterich
303 Dearborn Hall
Department of Computer Science
Oregon State University
Corvallis, OR 97331
tgd@cs.orst.edu
Principal investigators:
Thomas G. Dietterich, Department of Computer Science
Ron Nielson, US Forest Service
OSU is an Affirmative Action/Equal Opportunity Employer and Complies
with Section 504 of the Rehabilitation Act of 1973. OSU has a policy
of being responsive to the needs of dual-career couples.
Closing Date: July 5, 1994
------------------------------
Subject: Research position available
From: Jakob Carlstr|m <jakobc@Mordred.DoCS.UU.SE>
Date: Mon, 13 Jun 1994 10:59:34 +0200
RESEARCH POSITION AVAILABLE:
Hardware implementation of artificial neural networks
A research position is available in the field of hardware implementation
of artificial neural networks, at the Department of Computer Systems, Uppsala
University, Sweden.
The position is open to a scientist with a solid background in neural
networks and familiarity with analog and digital electronic circuit
construction as well as VLSI design. A candidate of postdoctoral or
equivalent status will be preferred.
The position will be in a new project for developing neural network
algorithms and hardware, aiming at VLSI implementations. The researcher
is expected to play a major role in this project. The position is tenable
for one year at minimum, possibly longer.
Uppsala University is Scandinavia's oldest university, founded in 1477, and
offers a stimulating research environment. The Department of Computer Systems
conducts research on artificial neural networks, real-time systems and
formal methods for concurrent systems.
The neural networks group was formed in 1991, and consists of four graduate
students supervised by Associate Professor Lars Asplund. We have published
reports on neural network-based control of digital telecom networks, and on
hardware architectures for neural networks.
Further information may be obtained from Associate Professor Lars Asplund,
Department of Computer Systems, Uppsala University, Box 325, S-751 05
Uppsala, Sweden; fax +46 18 55 02 25; email asplund@docs.uu.se.
Applicants should send a full CV, sample papers and the names and addresses
of two professional referees to the above address.
Closing date: August 8, 1994.
------------------------------
Subject: PostDoc position
From: Gilbert Chauvet <chauvet@bmsr14.usc.edu>
Date: Mon, 13 Jun 1994 17:41:01 -0700
Post Doc position available in: MATHEMATICAL BIOLOGY
Institute of Theoretical Biology, University of ANGERS (FRANCE),
beginning the 1st January 1995, for 2 years.
Qualification: PhD in Applied Mathematics
Project: Modeling in Neurobiology using non-local reaction diffusion
equations in general (numerical and theoretical aspects). Methods
will be applied to hippocampus.
Contact: Pr G.A. Chauvet, IBT
e-mail: chauvetg@ibt.univ-angers.fr
Phone: (33) 41 72 34 27
Fax: (33) 41 72 34 46
------------------------------
Subject: Fall School on Connectionism and Neural Nets HeKoNN 94 (in German)
From: franz@neuro.informatik.uni-ulm.de (Franz Kurfess)
Date: Mon, 30 May 1994 17:34:32 +0200
[[ Editor's Note: This may be for a limited audience, but we have many
subsribers in German-speaking countries. Further, other researchers with
good language skills may be interested. I readers feel strongly about
keeping the Digest English-only, please feel free to send comments to
neuron-request@psych.upenn.edu. -PM ]]
Below please find the announcement and call for participation of HeKoNN
94, a fall school on connectionism and neural networks to take place
October 10-14, 1994 near Muenster, Germany. The courses will be held in
German, so this will not be of much interest for people who don't speak
German.
Franz Kurfess
HeKoNN 94
Herbstschule Konnektionismus und Neuronale Netze
Muenster, 10.-14.10.1994
Im kommenden Oktober veranstalten die Fachgruppen "Konnektionismus" und
"Neuronale Netze" der GI (Gesellschaft fuer Informatik) eine Herbstschule
zu dem Themenbereich Konnektionismus und neuronale Netze. Sie bietet
Einfuehrungen und vertiefende Darstellungen zu folgenden Themen:
Grundlagen und Statistik, Implementierungen und Anwendungen, symbolischer
Konnektionismus und kognitiver Konnektionismus.
Konnektionistische Modelle und neuronale Netze sind einerseits inspiriert
von biologischen Vorbildern, insbesondere dem menschlichen Gehirn, dienen
andererseits inzwischen aber auch als praktikable Mechanismen zur Loesung
konkreter Probleme. Durch diese Dichotomie ergibt sich die Gefahr von
vielerlei Missverstaendnissen und unrealistischen Erwartungen, sei es
bezueglich der Leistungsfaehigkeit im Vergleich zu herkoemmlichen
Methoden, oder der Moeglichkeit biologische Systeme "nachzubauen".
Seit etwa Mitte der achtziger Jahre ist das Backpropagation Modell vielen
gelaeufig. Viel zu wenig bekannt sind aber theoretische Ergebnisse zu den
Eigenschaften dieses und alternativer Verfahren und ueber die
Zuverlaessigkeit der Methoden. Es ist heute klar, dass neuronale Netze in
enger thematischer Verwandtschaft mit Statistik, Funktionsapproximation,
und theoretischer Physik stehen und viele der dort gewonnen Erkenntnisse
auch hier anwendbar sind.
Darueberhinaus besteht noch ein recht grosses Defizit bei Fragen, die
sich mit den biologischen, kognitiven und psychologischen Aspekten
neuronaler Netze befassen. Hierbei dreht es sich um Konzepte zur
Modellierung von Verhaltensweisen und Denkprozessen auf der Basis
neuronaler Netze. Beispiele hierfuer sind die Repraesentation von
"Wissen", insbesondere die Verankerung von internen Darstellungen mit den
zugehoerigen Objekten der realen Welt, oder auch das Durchfuehren von
einfachen Schlussfolgerungen. Solche Fragen sind nicht nur von
akademischem Interesse, sondern ergeben sich auch beim Zusammenspiel von
eher symbolorientierter Wissensverarbeitung, z.B. in Expertensystemen,
und eher datenorientierten Verfahren etwa in der Mustererkennung. Und
genau an dieser Stelle liegen auch viele Schwierigkeiten von
herkoemmlichen Verfahren der Kuenstlichen Intelligenz, etwa in den
Bereichen Sprach- oder Bildverarbeitung.
Die Herbstschule bietet eine umfassende und fachuebergreifende
Darstellung des Themengebiets mit besonderer Betonung der obigen
Fragestellungen. Zwanzig aktive Wissenschaftler konnten als Dozenten
gewonnen werden, die jeweils 8-stuendige Kurse abhalten. Die Kurse sind
in vier Bereiche eingeteilt:
* Grundlagen und Statistik
* Implementierungen und Anwendungen
* Symbolischer Konnektionismus
* Kognitiver Konnektionismus
Im ersten Bereich -- Grundlagen und Statistik -- gibt es fuenf Kurse, in
denen die Grundlagen von Konnektionismus und neuronalen Netzen erlaeutert
werden. Dabei geht es zunaechst um die Vorstellung der haeufig
verwendeten Modelle sowie die Einfuehrung der entsprechenden Fachbegriffe
und Algorithmen. Ein Kurs bietete eine Einfuehrung in die Grundlagen,
Zielsetzungen und Forschungsfragen des Gebietes 'Konnektionismus' bzw.
'Neuronale Netzwerke' vom Standpunkt der K"unstlichen-Intelligenz-
Forschung. Ein zweiter Kurs diskutiert neuronale Netze aus dem
Blickwinkel der Approximationstheorie und Statistik. Hierbei wird
besonderes Gewicht auf die Diskussion der Eigenschaften der verschiedenen
Lern- und Optimierungsverfahren fuer die gaengigen Netzwerktypen gelegt.
Ein anderer Kurs untersucht die Eignung kuenstlicher Neuronaler Netze als
Modelle biologischer Vorgaenge. Die Betonung liegt hierbei auf dem
erforderlichen Realitaetsgrad der Netze bei der Simulation der
Informationsverarbeitung in den Nervensystemen von Lebewesen.
Entscheidend fuer den praktischen Einsatz neuronaler Netze bei der
Prognose und Prozesssteuerung ist die Zuverlaessigkeit der Ergebnisse.
Ein spezieller Kurs zu dieser Fragestellung stellt Methoden zur
Abschaetzung der Zuverlaessigkeit, zur Verbesserung der
Prognosegenauigkeit und zur Optimierung der Netzwerktopologie vor.
Besonders vielversprechend sind hier genetische Algorithmen, die ein
globales Optimum fuer Funktionen mit vielen Nebenmaxima bestimmen
koennen. Diese Verfahren werden in einem weitereren Kurs diskut iert.
Der zweite Bereich -- Implementierungen und Anwendungen -- beinhaltet
vier Kurse. In diesen Kursen werden zum einen mit SNNS und SESAME zwei
weit verbreitete Public-Domain-Simulationswerkzeuge mit graphischer
Benutzeroberflaeche vorgestellt, wobei praktische Uebungen am Rechner
vorgesehen sind. Ein Schwerpunkt dieser Simulatoren ist die flexible
Aenderung bestehender Netzstrukturen sowie die schnelle und sichere
Modifikation von Lernverfahren. Waehrend der SNNS grossen Wert auf die
graphischen Oberfl"ache unter X-Windows zur Generierung, Visualisierung
und Modifikation der neuronalen Netze legt, hat Sesame seinen Schwerpunkt
in einem modularen Experimentdesign, dass schnellen flexiblen Austausch
einzelner Komponenten sicherstellt und durch Modularisierung die
Konstruktion neuer Algorithmen unterstuetzt. Ein anderer Kurs
praesentiert den derzeitigen Stand der Technik bei der
Hardwareimplementierung neuronaler Netze. Hierbei handelt es sich
einerseits um Zusatzkarten f"ur konventionelle Arbeitsplatzrechner und
spezielle Parallelrechnersysteme, und zum anderen um Architekturen auf
der Basis von anwendungsspezifischen mikroelektronischen Bausteinen, die
entweder digital oder analog sein k"onnen. Als eine Anwendung werden
schliesslich neuronale Netze im Bereich der Robotik diskutiert.
Kuenstliche Neuronale Netze erscheinen hier geignet weil sie aus
Trainingsbeispielen selbst"andig relevante Informationen extrahieren
koennen. Schwerpunkt ist die Frage, ob sich hieraus in der Praxis
Vorteile gegenueber den klassischen analytischen Verfahren der Robotik
ergeben.
Der dritte Bereich -- Symbolischer Konnektionismus -- versucht, den
Zusammenhang herzustellen zwischen symbolorientierten Methoden der
Kuenstlichen Intelligenz und sub-symbolischen Methoden, die meist im
sensornahen Bereich, also bei der Datenerfassung vorliegen. Neuronale
Netze werden gerade im datennahen Bereich oft mit Erfolg eingesetzt, sind
jedoch nicht so ohne weiteres dafuer geeignet, Manipulationen auf
Zeichenreihen von Symbolen durchzufuehren. Eine wichtige Fragestellungen
hierbei ist die Ueberfuehrung von Rohdaten, wie etwa von einer Kamera
aufgenommene Bilder oder von einem Mikrophon registrierte akustische
Signale, in eine symbolische Form, auf der dann konventionelle Werkzeuge
wie Expertensysteme aufsetzen koennen. Ein anderer wichtiger Aspekt ist
die Extraktion von Wissen aus neuronalen Netzen, etwa zur Erklaerung
ihres Verhaltens oder zur Darstellung der von dem Netzwerk gelernten
Information in Form von Regeln.
Der letzte Bereich schliesslich -- Kognitiver Konnektionismus -- befasst
sich mit der Verwendung neuronaler Netze als Modelle fuer Wahrnehmung und
Denkprozesse. Zum einen werden hier wichtige grundlegende Probleme
dieser Modelle vor einem eher philosophischen Hintergrund diskutiert, zum
anderen aber auch Ansaetze zur Modellierung von Phaenomenen wie
Konzeptrepraesentation, Lernen, und Gedaechtnis besprochen. Weitere
Themen betreffen die Untersuchung und Modellierung von
informationsverarbeitenden Teilsystemen im Gehirn, etwa der
Sprachverarbeitung oder des visuellen Systems.
Die Kurse zu den vier obigen Bereichen werden parallel abgehalten; es ist
hierbei jedoch nicht notwendig, einen Bereich als Gesamtes auszuwaehlen,
sondern die Teilnehmer koennen und sollen Kurse aus verschiedenen
Bereichen belegen. Die Herbstschule wird im Jugendgaestehaus Aasee bei
Muenster stattfinden, wo sowohl die Teilnehmer als auch die Dozenten
untergebracht sein werden. Dadurch soll die Moeglichkeit geboten werden,
auch ausserhalb der eigentlichen Lehrveranstaltungen in einer zwanglosen
Atmosphaere ueber interessante Fragestellungen weiterzudiskutieren.
Angesprochen werden sollen insbesondere fortgeschrittenne Studenten sowie
Praktiker aus Forschung und Industrie. Die Zahl der Teilnehmer ist auf
100 beschraenkt. Der Preis fuer Studenenten wurde relativ niedrig
gehalten (ca. 410,- DM incl Tagungsunterlagen und Vollpension). Um das
Niveau der Tagung zu sichern, erfolgt die Auswahl der Teilnehmer auf
Grund einer Bewerbung. Hierbei werden Vorkenntnisse, praktische
Erfahrungen und das spezielle Interesse an Fragen des Konnektionismus und
der Neuronalen Netze beruecksichtigt. Anmeldeschluss ist der
- - - 1. Juli 1994 - - -
Im Organisations- und Programmkomitee sind
Ingo Duwe, Uni Bielefeld,
Franz Kurfess, Uni Ulm,
Gerhard Paass, GMD Sankt Augustin (Vorsitz),
Guenther Palm, Uni Ulm,
Helge Ritter, Uni Bielefeld,
Stefan Vogel, Uni Koeln.
Weitere Informationen sind erhaeltlich per anonymem ftp von
"ftp.gmd.de", Directory "/Learning/neural/HeKoNN94", per electronic
mail von "hekonn@neuro.informatik.uni-ulm.de",
oder vom
Tagungssekretariat HeKoNN 94
c/o Birgit Lonsinger,
Universitaet Ulm
Fakultaet fuer Informatik
Abteilung Neuroinformatik
D-89069 Ulm
Tel: 0731 502 4151
Fax: 0731 502 4156
------------------------------
End of Neuron Digest [Volume 13 Issue 33]
*****************************************
