Copy Link
Add to Bookmark
Report
Neuron Digest Volume 02 Number 12
NEURON Digest Fri May 8 08:56:02 CDT 1987 Volume 2 / Issue 12
Today's Topics:
Wanted: Pattern Recognition Application
ICNN conference
Abstract of talk at GTE Labs -- statistical/numerical approach
III Seminar, 11am Wed, 6-May-87
1st Annual ICNN ***Meeting Update***
Conference - Philosophy/Psychology Conference
Seminar: PDP: Implications for Psychology and Neurobiology
Technical Report -- Connectionist model of classical conditioning
----------------------------------------------------------------------
Date: Wed 6 May 87 11:33:42-PDT
From: Ken Laws <Laws@stripe.sri.com>
Subject: Wanted: Pattern Recognition Application
I'm trying to work up a proposal for research in classificatory
neural networks. The computational end is easy, but I need an
application that will provide the data for experimentation. I
need either real-world data or (perhaps even better) a way to
synthesize interesting data. I'm already aware of some work in
image analysis, speech recognition, and character recognition,
but would be interested to hear from people who have current
problems that are not being adequately addressed. Can you suggest
any other pattern recognition problem or "signal" of particular
interest to the Air Force, Navy, Army, NASA, NSF, or other funding
agency?
-- Ken Laws
LAWS@STRIPE.SRI.COM
------------------------------
Date: 30 Apr 87 13:49:00 EST
From: "NRL::SZU" <szu%nrl.decnet@nrl3.arpa>
Subject: ICNN conference
PRESS RELEASE
-----------------
In response to intense international interest in neural network
modeling and its applications, the International Neural Network Society
(INNS) has been founded as a nonprofit public benefit corporation in
the state of California. Its purpose is to create a scientific and
educational forum for students, scientists, engineers, and others to
learn about and advance the state of knowledge concerning the modeling
of behavioral and brain processes, and the application of neural
modeling ideas to problems in technology.
The founding governing board of the Society includes Shun-ichi
Amari, James Anderson, Gail Carpenter, Walter Freeman, K. Fukushima, Lee
Giles, Stephen Grossberg, Robert Hecht-Nielsen, Teuvo Kohonen, Christoph
von der Malsburg, Carver Mead, Demetri Psaltis, David Rumelhart, G.
Sperling, T. Sejnowski, Harold Szu, and Bernard Widrow.
The official journal of the Society is called Neural Networks. The
editors-in-chief of the journal are Shun-ichi Amari, Stephen Grossberg,
and Teuvo Kohonen. Pergamon Press is the publisher. The large and
distinguished editorial board includes experts in all aspects of the
field who will have full authority over editorial decisions concerning
the articles they handle, and will play an active role in organizing
such projects as special issues and invited review articles. In this
way, the journal will serve as a highly visible, cooperative,
interdisciplinary vehicle to bring together contributions from
psychologists, neurobiologists, computer scientists, mathematicians,
engineers, and physicists, and to enable this remarkably large and broad
group of scientists to easily find the full spectrum of high quality
quantitative models in this area in one place.
Applications for membership in the Society, whose dues ($45, $35 for
students) include a subscription to the journal, will be available at the
First International Conference on Neural Networks in San Diego on June
21-24, 1987. Inquires concerning the Society and membership can be
directed to Dr. Harold Szu, Naval Research Lab, Code 5756, Washington,
DC 20375-5000. Inquiries about the journal should be sent to the
editors-in-chief:
Prof. Shun-ichi Amari
University of Tokyo
Faculty of Engineering
Instrumentation Physics
Bunkyo-ku, Tokyo 113
JAPAN
Prof. Stephen Grossberg
Boston University
Ctr. for Adaptive Systems
Mathematics Department
111 Cummington Street
Boston, MA 02215
Prof. Teuvo Kohonen
Helsinki University of Technology
Technical Physics Dept.
Rakentajanaukio 2C
SF-02150 Espoo 15
FINLAND
Public Contest for INNS Seal
--------------------------------
A design for the INNS seal is publically solicited for the
Society. The winner will be selected during the board meeting on June
20, 1987. All entries should be submitted as follows:
- Format: in black-white only on 8X11 paper
- Deadline: No later than June 20, 1987. All entries should be sent to
IEEE First Annual International Conference on Neural Network
Sheraton Harbor Island East
1380 Harbor Island Drive
San Diego, CA 92101
ATTN: Dr. Harold Szu
- Prize: Public Recognition during the Banquet Reception of the San Diego
Conference
------------------------------
Date: Fri, 1 May 87 17:16:54 EDT
From: Rich Sutton <rich@gte-labs.csnet>
Subject: Abstract of talk at GTE Labs -- statistical/numerical approach
Abstract of talk at GTE Labs:
A STATISTICAL/NUMERICAL APPROACH TO DESIGNING ADAPTIVE INTELLIGENCE
Paul J. Werbos
Energy Information Administration
Friday, May 8, 10am
Room 3-131, GTE Labs
This talk will discuss a new approach towards building a general theory
of adaptive intelligent systems recently published in the IEEE
transactions. Starting from a classical looking definition of an
intelligent system, it argues that generalized methods for optimizing
over time must be the core of such a system, which narrows down the
design possibilities considerably. It provides a familiy of three
possible optimization methods, one of which (HDP) matches the few
(though highly successful) adaptive critic systems from AI, and another
(GDHP) which appears to offer maximum power as well as a match to the
gross structure of the mammalian brain. To complete the design of such
an intelligent system (based on GDHP), a general subsystem to develop
stochastic time-series models is also required; This can be done in a
variety of ways based on current statistical theory, which includes but
goes well beyond conventional perceptrons and content-addressable
memories. This work begins to suggest a taxonomy of designs, from the
simplest through to the most powerful, along with possibilities for
variation in certain arbitrary parameters which feel similar to the
taxonomy and variablity of natural intelligence.
-----------------------------------------------------------------------
To attend the talk, please arrive 15 minutes early, sign in, and wait in
the visitors lobby. Someone will come to the lobby shortly before the
talk to escort you to room 3-131. Reply to this message for directions
to GTE or to let us know you are coming. Your contact person is Rich
Sutton, rich@gte-labs.csnet, (617)466-4133.
------------------------------
Date: 1 May 87 15:02:48 EDT
From: LANTZ@RED.RUTGERS.EDU
Subject: III Seminar, 11am Wed, 6-May-87
The III, an Informal Seminar by and for Students
Title: Self-Organization in Neural Networks, with an Application
to Processor Assignment on Multi-Processor Networks.
Date: 6-May-87, Wednesday
Time: 11:00 AM
Place: Hill 423
Speaker: Brian Lantz
Keywords: Self-Organization, Neural Modelling, Processor Assignment
Parallel Distributed Processing, Great Stuff.
Abstract:
Neurons from the retina are mapped onto the cortex in an orderly
fashion, and neurons from the ears produce a perfect tonotopic mapping
on the auditory cortex. Models of self-organization may explain how
mappings from one space to another are possible without the need for
an explicit description of the mappings. Several very interesting
models of neural self-organization have been proposed, some with
astounding results.
The general problem of mapping one space onto another is found in the
processor assignment problem. Encapsulated code modules and their
communications paths are one space, and the physical processing nodes and
their interconnections are a second space. Mapping the communications graph
onto the physical processor array is the processor assignment problem.
I propose the use of self-organization techniques to induce communicating
code modules to assume their optimal locations within the processing
network.
The mapping technique can map from any M-dimensional space to any
N-dimensional space (or any subset, which need not be regular
or contiguous.) I will also touch on a hopeful application of fractals
to the processor assignment problem.
Right before the coffee hour: Great Stuff.
------------------------------
Date: 5-MAY-1987 11:23
From: MIKE%BUCASA.BITNET@WISCVM.WISC.EDU
Subject: 1st Annual ICNN ***MEETING UPDATE***
***MEETING UPDATE***
First Annual International Conference on Neural Networks
San Diego, California
June 21--24, 1987
There will be 16 scientific sessions with invited and contributed lectures, 5
plenary talks, continuous poster presentations, and many exhibits. Over 300
scientific presentations will be made. Twelve tutorials will be offered on June
20.
For information about registration, contact Nomi Feldman, Conference
Coordinator, 3770 Tansy Street, San Diego, CA 92121; (619) 453-6222.
The program of scientific sessions and their chairpersons are listed below.
------------------------------------------------------------------------------
*** JUNE 21, 1987 ***
(8:00 AM--11:30 AM)
Self-Organization (Grossberg, Kohonen)
Cooperative and Competitive Network Dynamics (Hirsch)
(11:30 AM--1:00 PM)
LUNCH
(1:00 PM--4:30 PM)
Knowledge Processing (Kelso)
Learning Algorithms, I (Anderson, Widrow)
(4:30 PM--5:15 PM)
BREAK
(5:15 PM--7:30 PM)
PLENARY TALKS
Stephen Grossberg
Teuvo Kohonen
-----
*** JUNE 22, 1987 ***
(8:00 AM--11:30 AM)
Network Architectures, I (Cohen)
Learning Algorithms, II (Rumelhart, Zipser)
(11:30 AM--1:00 PM)
LUNCH
(1:00 PM--4:30 PM)
Adaptive Resonance (Carpenter)
Network Architectures, II (Amari)
(4:30 PM--5:15 PM)
BREAK
(5:15 PM--7:30 PM)
PLENARY TALKS
Carver Mead
John Hopfield
-----
*** JUNE 23, 1987 ***
(8:00 AM--11:30 AM)
Vision (Fukushima, Mingolla)
Neurobiological Connections (Levine, Sperling)
(11:30 AM--1:00 PM)
LUNCH
(1:00 PM--4:30 PM)
Speech Recognition and Synthesis (Elman, Stork)
Electrical Neurocomputers (Hecht-Nielsen, Penz)
(4:30 PM--5:15 PM)
BREAK
(5:15 PM--6:30 PM)
PLENARY TALK
Bernard Widrow
RECEPTION
-----
*** JUNE 24, 1987 ***
(8:00 AM--11:30 AM)
Combinatorial Optimization (Szu)
Optical Neurocomputers (Goodman, Guest)
(11:30 AM--1:00 PM)
LUNCH
(1:00 PM--4:30 PM)
Robotics (Bavarian, Stubberud)
Novel Applications (Giles, Kellum)
-----
Please Post this Message to Your Mailing List
Thanks much for your troubles.
Replies to:
mike%bucasa.bitnet@wiscvm.wisc.edu (arpanet)
mike@bucasa (bitnet)
------------------------------
Date: 4 Apr 87 05:55:03 GMT
From: princeton!mind!harnad@RUTGERS.EDU (Stevan Harnad)
Subject: Conference - Philosophy/Psychology Conference
[Portions of an excerpt from AIList Digest - MTG]
Program of the 13th Annual Meeting of the Society for Philosophy and Psychology
June 21 -23, University of California, San Diego
For program information: William Bechtel (SPP Program Chairman),
Philosophy Department, Georgia State University, Atlanta GA 30303-3083
phone: (404)-658-2277 bitnet address: psuvax1!phlpwb%GSUMVS1.BITNET
For membership information: Patricia Kitcher, Philosophy Department,
University of California-San Diego, La Jolla CA 92093
arpanet address: sdcsvax!ir205%sdcc6
-------- SUNDAY, JUNE 21, 1987 --------
SESSION II: Computational Theories of Mind
Chair: Owen Flanagan, Philosophy, Wellesley
Speaker: David Kirsch, Artificial Intelligence, MIT
"The Concept of Computation in Connectionist Systems"
Commentator: Brian Cantwell Smith, Computer Science, Xerox PARC
Speaker: Joseph Levine, Philosophy, North Carolina State University
"Demonstrative Thought"
Commentator: La Verne Shelton, Educational Testing Service, Princeton
3:30-5:00pm INVITED LECTURE: LANGUAGES OF THE DEAF
Chair: Adele Abrahamsen, Language Research Center, Georgia State
Speaker: Howard Poizner, Salk Institute, San Diego
"Brain Function for Language: Perspectives from Another Modality"
----- MONDAY, JUNE 22, 1987 --------
9:00-11:30am SYMPOSIUM: CONNECTIONISM AND IMAGE SCHEMATIC STRUCTURES
Chair: Patricia Churchland, Philosophy, University California, San Diego
Speakers: David Rumelhart, Psychology, University of California, San Diego
George Lakoff, Linguistics, University of California, Berkeley
Mark Johnson, Philosophy, Southern Illinois University
Terrence Sejnowski, Biophysics, Johns Hopkins University
----- TUESDAY, JUNE 23, 1987 --------
11:15-12:30pm INVITED LECTURE: Memory and Brain
Chair:
Speaker: Larry R. Squire, Psychiatry, University of California, San Diego
"Memory and Brain: Neural Systems and Behavior"
--
Stevan Harnad (609) - 921 7771
{bellcore, psuvax1, seismo, rutgers, packard} !princeton!mind!harnad
harnad%mind@princeton.csnet harnad@princeton.ARPA harnad@mind.Princeton.EDU
------------------------------
Date: 06 May 87 15:17:28 bst
From: N.STROUD%UK.AC.EDINBURGH@ac.uk
Subject: Seminar: PDP: Implications for Psychology and Neurobiology
PARALLEL DISTRIBUTED PROCESSING
Implications for Psychology and Neurobiology
Department of Experimental Psychology
University of Oxford - England
1st July 1987
This symposium forms part of the Experimental Psychology Society's annual
Summer Meeting, which this year is being held jointly with the Experimental
Division of the Canadian Psychological Association. It is hoped that the
symposium will attract people in both academic life and industry. Information
about the symposium, speakers and registration is given below.
Recent years have witnessed a substantial resurgence of interest in the idea
of "network", "parallel" or "distributed" memory systems (eg Hinton and
Anderson, 1981, Rumelhart and McClelland, 1986). In considering the themes
for the proposed symposium, the Society decided to invite recognised experts
from both cognitive psychology and the neurosciences. We shall retain some
separation between these two themes by devoting the morning session to more
cognitive issues, and the afternoon to questions of implementation at the
physiological and biochemical level. The morning session will give particular
attention to the recent work of Rumelhart and McClelland (1985,1986) concerning
the representation of "general and specific information", for example the
extraction of prototypes; and to the work of Rumelhart, Hinton and Williams
(1986) on new algorithms for parallel distributed processing. The afternoon
will concentrate on the role of synaptic plasticity in conditioning and
information storage, from mollusc (Kandel and Hawkins, 1984) to mammal (Lynch
and Baudry, 1984; Lynch, 1986). In order to have an informed discussion, we
are particularly keen that the implications of these ideas for psychology and
neurobiology are brought to the fore - and have therefore included certain
speakers whose primary experience is in experimental work on concept-learning,
the neuropsychology of language disorders, language and perception, single-unit
recording and long-term potentiation. Certain criticisms of the PDP approach
may also be brought out into the open. For example, Rumelhart and
McClelland's (1985) approach has been criticised on the grounds of mixing
different "levels of explanation" (Broadbent, 1985). Kandel and Hawkins'
(1984) intriguing idea that "sensitisation" might be part of the "alphabet"
of associative conditioning is not without its problems.
Oxford University Press will publish the proceedings of the symposium.
* *
References:
Broadbent,D.E. (1985) "A Question of Levels: Comment on McClelland and
Rumelhart", Journal of Experimental Psychology: General, 114, 189-192.
Hinton,G.E. and Anderson,J.A. (1981) "Parallel models of associative memory",
Erlbaum, Hillsdale, New Jersey.
Kandel,E.R. and Hawkins,R.D. (1984) "Is there a Cell-Biological Alphabet for
simple forms of learning?", Psychological Review, 91, 375-391.
Lynch,G.S. and Baudry,M. (1984) "The biochemistry of memory: A new and
specific hypothesis", Science, 224, 1057-1063.
Lynch,G.S. (1986) "Synapses, Circuits and the Beginnings of Memory", MIT Press,
Cambridge, MA, in press.
McClelland,J.L. and Rumelhart,D.E. (1985) "Distributed memory and the
representation of general and specific information", Journal of
Experimental Psychology: General, 114, 158-198.
Rumelhart,D.E., McClelland,J.L. and the PDP Research Group (1986) "Parallel
Distributed Processing: Explorations in the Microstructure of Cognition",
vols 1 & 2, MIT Press, Cambridge, MA.
Rumelhart,D.E., Hinton,G.E. and Williams,R.J. (1986) "Learning representations
by back-propagating errors", Nature, 323, 533-536.
* *
PROGRAMME
9:00 Introduction by morning Chairman, Professor N.S.Sutherland (Centre for
Research on Perception and Cognition, University of Sussex).
9:05 D.E.Rumelhart and J.L.McClelland (University of California, San Diego,
and Carnegie-Mellon University): "Parallel Distributed Memory Systems".
10:00 B.W.A.Whittlesea (Dept of Psychology, Mount-Allison University, New
Brunswick): "The distributed representation of concepts".
10:30 M.Seidenberg, J.L.McClelland and K.Patterson (McGill University,
Carnegie-Mellon University, MRC Applied Psychology Unit, Cambridge):
"A distributed developmental model of visual word recognition and
pronunciation".
11:00 COFFEE
11:30 G.E.Hinton (Dept of Computer Science, Carnegie-Mellon University):
"Learning by modifying connection strengths in parallel networks".
12:00 S.Pinker (Dept of Brain and Cognitive Science, MIT): "PDP Systems and
the Human Language Faculty".
12:30 Open General Discussion with all morning speakers, chaired by
Professor Sutherland.
1:00 LUNCH (sandwiches in foyer)
2:00 Introduction by afternoon Chairman, Dr D.Willshaw (MRC Neural Development
Unit, University of Edinburgh).
2:05 R.D.Hawkins and E.R.Kandel (Center for Neurobiology and Behaviour,
Columbia University): "A Cell-Biological Alphabet for associative
learning".
2:45 N.J.Mackintosh (Dept of Experimental Psychology, University of
Cambridge): "Associative learning and Parallel Distributed Processing".
3:15 E.T.Rolls (Dept of Experimental Psychology, University of Oxford):
"Neuronal networks in the Primate and Stimulus Categorisation".
3:45 TEA
4:15 R.G.M.Morris (Dept of Pharmacology, University of Edinburgh): "Does
synaptic plasticity play a role in information storage?".
4:45 G.S.Lynch (Centre for the Neurobiology of Learning and Memory,
University of California, Irvine): "Synapses, Circuits and the
Beginnings of Memory".
5:30 Open General Discussion with all speakers, chaired by Dr Willshaw.
* *
REGISTRATION
The meeting will be held at the Department of Experimental Psychology, South
Parks Road, University of Oxford, on 1st July 1987.
The Experimental Psychology Society does not normally require registration at
its meetings. Members of the Society and their guests will therefore be
admitted free of charge. However, members should write to the Hon Secretary
giving the names and addresses of any guests they wish to bring to the meeting,
so that programmes for the symposium may be distributed in advance. Admission
to the Lecture Theatre will be by programme.
Because of the cost of organising the meeting, the Society has decided with
regret to levy a registration charge on non-members: this will be seven pounds
and fifty pence for academics, and twenty pounds for non-academics.
Postgraduate students will be admitted free of charge (on proof of status).
You may register from 8:30am on the day of the meeting, or to receive the
program in advance, by writing to:
The Hon Secretary,
Experimental Psychology Society,
Psychological Laboratory,
University of St Andrews,
Fife,
Scotland, U.K.
KY16 9JU
(or tel 0334 76161 ext 7173, and ask for Mrs Anderson)
* * *
------------------------------
Date: Thu, 30 Apr 87 10:32:58 EDT
From: Rich Sutton <rich@gte-labs.csnet>
Subject: Technical Report -- Connectionist model of classical conditioning
A TEMPORAL-DIFFERENCE MODEL OF CLASSICAL CONDITIONING
Richard S. Sutton
GTE Laboratories Inc.
Andrew G. Barto
University of Massachusetts
ABSTRACT
Rescorla and Wagner's model of classical conditioning has been one of
the most influential and successful theories of this fundamental
learning process. The learning rule of their theory was first described
as a learning procedure for connectionist networks by Widrow and Hoff.
In this paper we propose a similar confluence of psychological and
engineering constraints. Sutton has recently argued that adaptive
prediction methods called {\it temporal-difference methods\/} have
advantages over other prediction methods for certain types of problems.
Here we argue that temporal-difference methods can provide detailed
accounts of aspects of classical conditioning behavior. We present a
model of classical conditioning behavior that takes the form of a
temporal-difference prediction method. We argue that it is an
improvement over the Rescorla-Wagner model in its handling of
within-trial temporal effects such as the ISI dependency, primacy
effects, and the facilitation of remote associations in serial-compound
conditioning. The new model is closely related to the model of
classical conditioning that we proposed in 1981, but avoids some of the
problems with that model recently identified by Moore et al. We suggest
that the theory of adaptive prediction on which our model is based
provides insight into the functionality of classical conditioning
behavior.
--------------------------------------------------------------------------
This paper has been submitted to the 1987 Cognitive Science conference.
It is currently available as a technical report; reply to
rich@gte-labs.CSNet or R. Sutton, GTE Labs, Waltham, MA, 02254.
------------------------------
End of NEURON-Digest
********************
