Copy Link
Add to Bookmark
Report
Neuron Digest Volume 06 Number 05
Neuron Digest Tuesday, 23 Jan 1990 Volume 6 : Issue 5
Today's Topics:
Announcement: 1990 Connectionist Models Summer School
Call for papers for GLOBECOM '90
Artificial Life Workshop
neural nets and light adaptation (TR)
MIT AI Lab memo 1164
Reprint Available
Technical Report available
two more Tech Reports available
Behavioral Neuroscience Faculty Position at Boulder
SAB90 Call for Papers
Send submissions, questions, address maintenance and requests for old issues to
"neuron-request@hplabs.hp.com" or "{any backbone,uunet}!hplabs!neuron-request"
Use "ftp" to get old issues from hplpm.hpl.hp.com (15.255.176.205).
------------------------------------------------------------
Subject: Announcement: 1990 Connectionist Models Summer School
From: elman@amos.ling.ucsd.edu (Jeff Elman)
Organization: Univ. of Calif., San Diego
Date: 20 Dec 89 19:32:25 +0000
December 20, 1989
ANNOUNCEMENT & SOLICITATION FOR APPLICATIONS
CONNECTIONIST MODELS SUMMER SCHOOL / SUMMER 1990
UCSD
La Jolla, California
The next Connectionist Models Summer School will be
held at the University of California, San Diego from June 19
to 29, 1990. This will be the third session in the series
which was held at Carnegie Mellon in the summers of 1986 and
1988. Previous summer schools have been extremely success-
ful, and we look forward to the 1990 session with anticipa-
tion of another exciting summer school.
The summer school will offer courses in a variety of
areas of connectionist modelling, with emphasis on computa-
tional neuroscience, cognitive models, and hardware imple-
mentation. A variety of leaders in the field will serve as
Visiting Faculty (the list of invited faculty appears
below). In addition to daily lectures, there will be a
series of shorter tutorials and public colloquia. Proceed-
ings of the summer school will be published the following
fall by Morgan-Kaufmann (previous proceedings appeared as
'Proceedings of the 1988 Connectionist Models Summer
School', Ed., David Touretzky, Morgan-Kaufmann).
As in the past, participation will be limited to gradu-
ate students enrolled in PhD. programs (full- or part-time).
Admission will be on a competitive basis. Tuition is sub-
sidized for all students and scholarships are available to
cover housing costs ($250).
Applications should include the following:
(1) A statement of purpose, explaining major areas of
interest and prior background in connectionist model-
ing (if any).
(2) A description of a problem area you are interested in
modeling.
(3) A list of relevant coursework, with instructors'
names and grades.
(4) Names of the three individuals whom you will be ask-
ing for letters of recommendation (see below).
(5) If you are requesting support for housing, please
include a statement explaining the basis for need.
Please also arrange to have letters of recommendation
sent directly from three individuals who know your current
work.
Applications should be sent to
Marilee Bateman
Institute for Neural Computation, B-047
University of California, San Diego
La Jolla, CA 92093
(619) 534-7880
All application material must be received by March 15, 1990.
Decisions about acceptance and scholarship awards will be
announced April 1. If you have further questions, contact
Marilee Bateman (address above), or one of the members of
the Organizing Committee.
Jeff Elman Terry Sejnowski
UCSD UCSD/SalkInstitute
elman@amos.ucsd.edu terry@sdbio2.ucsd.edu
Geoff Hinton Dave Touretzky
Toronto CMU
hinton@ai.toronto.edu touretzky@cs.cmu.edu
------------
INVITED FACULTY:
Yaser Abu-Mostafa (CalTech) Richard Lippmann (MIT Lincoln Labs)
Dana Ballard (Rochester) James L. McClelland (Carnegie Mellon)
Andy Barto (UMass/Amherst) Carver Mead (CalTech)
Gail Carpenter (BU) David Rumelhart (Stanford)
Patricia Churchland (UCSD) Terry Sejnowski (UCSD/Salk)
Jack Cowan (Chicago) Al Selverston (UCSD)
Jeff Elman (UCSD) Paul Smolensky (Colorado)
Jerry Feldman (ICSI/UCB) David Tank (Bell Labs)
Geoffrey Hinton (Toronto) David Touretzky (Carnegie Mellon)
Michael Jordan (MIT) Halbert White (UCSD)
Teuvo Kohonen (Helsinki) Ron Williams (Northeastern)
George Lakoff (UCB) David Zipser (UCSD)
INVITED DISCUSSION LEADERS:
Rik Belew (UCSD) Marty Sereno (UCSD
Gary Cottrell (UCSD) Shawn Lockery (UCSD
Richard Durbin (Stanford)
------------------------------
Subject: Call for papers for GLOBECOM '90
From: nirwan ansari fac ee <ang@hertz.njit.edu>
Date: Thu, 21 Dec 89 11:35:32 -0500
The 1990 IEEE Global Telecommunications Conference (GLOBECOM 90) will be
held in San Diego, California, Decemebr 2-5, 1990. I was asked by the
technical committee to organize a session "Neural Networks in
Communication Systems." You are cordially invited to submit an original
technical paper related to this topic for consideration for GLOBECOM 90.
The SCHEDULE is as follows:
Complete Manuscript Due 3/15/1990
Notification of Acceptance Mailed 5/30/1990
Camera-ready Manuscript Due 8/20/1990
INSTRUCTIONS:
The title page must include the author's name, complete return address,
telephone, telex and fax number and abstract (100 words). For papers with
multiple authors, please designate the author to whom all correspondence
should be sent by listing that author first. All other pages should have
the title and first author of the paper. The manuscript should not exceed
3,000 words in English. Page charges will be assessed for camera-ready
copies exceeding five pages. Please send six double-spaced copies of the
manuscript in English to:
Dr. Arne Mortensen
GLOBECOM '90
Technical Program Secretary
M/A-COM Government Systems
3033 Science Park Road
San Diego, CA 92121
Phone:(619) 457-2340
Telex:910-337-1277
FAX:(619) 457-0579,
and a copy to me:
Dr. Nirwan Ansari
GLOBECOM '90 Neural Network Session Organizer
Electrical and Computer Engineering Department
New Jersey Insitute of Technology
University Heights
Newark, NJ 07102
Phone:(201) 596-5739.
Please also indicate in your cover letter to Dr. Mortensen that you have
communicated with and sent me a copy of your manuscript for consideration
for the "Neural Networks in Communication Systems" Session.
For further questions, please feel free to contact me using the above
address or the e-mail address, ang@hertz.njit.edu (node address:
128.235.1.26).
------------------------------
Subject: Artificial Life Workshop
From: cgl@lanl.gov (C G Langton)
Organization: Los Alamos National Laboratory
Date: 21 Dec 89 23:13:06 +0000
FINAL ANNOUNCEMENT !!!!
ARTIFICIAL LIFE
---------------
A workshop on the synthesis of
living and evolving artifacts.
February 5-9, 1990
Santa Fe, New Mexico
Sponsored by
------------
The Center for Nonlinear Studies, LANL
and
The Santa Fe Institute
Self-Organizers
---------------
Doyne Farmer
Chris Langton
Steen Rasmussen
Charles Taylor
Artificial Life has only recently emerged as a coherent field of
scientific research. Its primary methodological approach is to study
life and evolution by attempting to actually create living and/or
evolving processes within computers, beakers, or other ``artificial''
media. Its primary goal is to abstract the ``logical form'' of life
from its material basis - and to construct a truly general theory of
living systems, one which will be capable of treating life wherever it
is found in the universe and whatever it is made of. ``Artificial'' Life
can contribute to the study of ``real'' life by helping to locate
life-as-we-know-it within the larger context of life-as-it-could-be,
in any of its possible incarnations.
This will be the second workshop on the topic of Artificial Life. The
workshop will include invited and contributed talks, demonstrations,
and discussions on the many scientific, technical, philosophical, and
moral issues surrounding the increasing attempts to synthesize life
artificially. We will also have an artificial ``4H show'' with prizes
for the best artificial life-forms.
Specific investigations in the field of Artificial Life include attempts
to synthesize, simulate, or otherwise recreate the following:
- the emergence of autocatalytic sets within soups of artificial polymers;
- the evolution of strings of code using Genetic Algorithms;
- self-reproducing bit-strings, clay-crystals, RNA molecules, or LEGO-robots;
- the emergence of cooperativity, colonial organization, multi-cellularity,
and hierarchical organization;
- the embryological processes of growth, development, and differentiation;
- the emergence of social behavior in populations of artificial insects;
- the emulation of population and ecosystem dynamics;
- the implementation of artificial environments, logical universes,
or ``virtual realities'' sufficiently rich to support the open-ended
evolution of embedded ``organisms'';
- cultural evolution, including the origin and evolution of socio-
cultural institutions, and the evolution of natural language in its
role as a vehicle for cultural inheritance;
- the dynamics of self-propagating information structures such as
biological and computer viruses;
Many of the investigations mentioned above will be reported on or
discussed at the workshop.
We expect that there will also be plenty of debate on the question of
whether or not symbolic processes within computers can be considered
``alive'' in principle, or whether they could be capable of participating
in anything like truly open-ended evolution. These debates will probably
parallel to a large extent the debates in the AI community on whether
processes within computers can considered to be ``intelligent'' or
``conscious.''
We are also encouraging presentations and/or debates on the moral and
social consequences of achieving the capability to create living things.
The mastery of the technology of life will easily overshadow any of our
previous technological accomplishments - even our mastery of the technology
of death - in terms of the burden of responsibility which it places on our
shoulders. As was the case for the mastery of atomic fission and fusion,
the potential abuses are directly proportional to the potential benefits.
Once again, we are in a position where our technical understanding of nature
is far in advance of our understanding of the potential consequences
of mastering or deploying the technology. This is not an enterprise to
be undertaken lightly, or to be pursued in the cause of such shortsighted
goals as fleeting military advantage.
The increasing spread and sophistication of computer viruses is evidence
both of the imminence of this new era in the history of life, and of the
complexity of the problems and issues that will be facing all of us in
the not-too-distant future.
We welcome your presence and contribution on any aspect of Artificial
Life that you consider worth presenting or discussing with others
who are interested in such issues. Whether you are a scientist, an
engineer, a philosopher, an artist, or just a concerned citizen, we
feel that ALL points of view need to be aired at this early stage in
the evolution of Artificial Life.
For further information and/or registration materials, contact:
Andi Sutherland
The Santa Fe Institute
1120 Canyon Rd.
Santa Fe, New Mexico
87501
505-984-8800
andi@sfi.santafe.edu
The deadline for contributions is Dec. 31, 1989. Registrations for
the workshop will be accepted right up to the date of the workshop.
Some limited financial assistance will be available for the truly
needy.
The proceedings of the first Artificial Life Workshop, held at
the Center for Nonlinear Studies, Los Alamos, New Mexico in 1987,
are available from Addison Wesley: "Artificial Life: The proceedings
of an interdisciplinary workshop on the synthesis and simulation
of living systems", edited by Christopher G. Langton, Volume #6
in Addison Wesley's `Santa Fe Institute Studies in the Sciences
of Complexity' series. They can be ordered toll free by calling
800-447-2226. The order codes are:
Hardback (about $40) ISBN 0-201-09346-4
Paperback (about $20) ISBN 0-201-09356-1
------------------------------
Subject: neural nets and light adaptation (TR)
From: "Dr. Josef Skrzypek" <skrzypek@CS.UCLA.EDU>
Date: Thu, 21 Dec 89 15:14:38 -0800
NEURAL NETWORK CONTRIBUTION TO LIGHT ADAPTATION:
FEEDBACK FROM HORIZONTAL CELLS TO CONES
JOSEF SKRZYPEK
Machine Perception Laboratory, Computer Science Department
and
CRUMP Institute of Medical Engineering.
UCLA
SUMMARY
Vertebrate cones respond to a stepwise increase in localized light
intensity with a graded potential change of corresponding amplitude.
This S-shaped intensity-response (I-R) relation is limited to 3 log units
of the stimulating light and yet, cone vision remains functional between
twilight and the brightest time of day. This is in part due to light
adaptation mechanism localized in the outer segment of a cone. The
phenomenon of light adaptation can be described as a resetting of the
system's response-generation mechanism to a new intensity domain that
reflects the ambient illumination. In this project we examined spatial
effects of annular illumination on resetting of I-R relation by measuring
intracellular photoresponses in cones. Our results suggest that
peripheral illumination contributes to the cellular mechanism of
adaptation. This is done by a neural network involving feedback synapse
from horizontal cell to cones. The effect is to unsaturate the membrane
potential of a fully hyperpolarized cone, by "instantaneously" shifting
cone's I-R curves along intensity axis to be in register with ambient
light level of the periphery. An equivalent electrical circuit with three
different transmembrane channels leakage, photocurrent and feedback was
used to model static behavior of a cone. SPICE simulation showed that
interactions between feedback synapse and the light sensitive conductance
in the outer segment can shift the I-R curves along the intensity domain,
provided that phototransduction mechanism is not saturated during
maximally hyperpolarized light response.
Key words: adaptation, feedback, cones, retina, lateral interactions
Josef Skrzypek
Computer Science Department
3532D Boelter Hall
UCLA Los Angeles, California 90024
INTERNET: SKRZYPEK@CS.UCLA.EDU
------------------------------
Subject: MIT AI Lab memo 1164
From: Tomaso Poggio <poggio@ai.mit.edu>
Date: Wed, 27 Dec 89 11:07:40 -0500
the following technical report is available from the MIT AI Lab
Publication Office (send e-mail to liz@ai.mit.edu)
Networks and the Best Approximation Property
by
Federico Girosi and Tomaso Poggio
ABSTRACT
Networks can be considered as approximation schemes. Multilayer networks
of the backpropagation type can approximate arbitrarily well continuous
functions (Cybenko, 1989; Funahashi, 1989; Stinchcombe and White, 1989).
We prove that networks derived from regularization theory and including
Radial Basis Functions (Poggio and Girosi, 1989, AI memo 1140), have a
similar property. From the point of view of approximation theory,
however, the property of approximating continuous functions arbitrarily
well is not sufficient for characterizing good approximation schemes.
More critical is the property of {\it best approximation}. The main
result of this paper is that multilayer networks, of the type used in
backpropagation, are not best approximation. For regularization networks
(in particular Radial Basis Function networks) we prove existence and
uniqueness of best approximation.
------------------------------
Subject: Reprint Available
From: Selma M Kaufman <smk@flash.bellcore.com>
Date: Thu, 28 Dec 89 11:16:57 -0500
Learning of Stable States in Stochastic Asymmetric Networks
Robert B. Allen and Joshua Alspector
Bellcore TR-AR-89-351
December, 1989
Boltzmann-based models with asymmetric connections are investigated.
Although they are initially unstable, we find that these networks
spontaneously self-stablize as a result of learning. Moreover, we find
that pairs of weights symmetrize during learning; however, the symmetry
is not enough to account for the observed stability. To characterize the
system we consider how its entropy is affected by learning and the
entropy of the information stream. Finally, the stability of an
asymmetric network was confirmed with an electronic model.
For paper copies, contact:
Selma Kaufman, Bellcore, 2M-356, 445 South St., Morristown, NJ 07960-1910.
smk@flash.bellcore.com
------------------------------
Subject: Technical Report available
From: Geoffrey Hinton <hinton@ai.toronto.edu>
Date: Thu, 28 Dec 89 13:15:13 -0500
Please do not reply to this message. To order a copy of the TR described
below, please send email to carol@ai.toronto.edu
_________________________________________________________________________
DETERMINISTIC BOLTZMANN LEARNING
IN NETWORKS WITH ASYMMETRIC CONNECTIVITY
Conrad C. Galland and Geoffrey E. Hinton
Department of Computer Science
University of Toronto
10 Kings College Road
Toronto M5S 1A4, Canada
Technical Report CRG-TR-89-6
The simplicity and locality of the "contrastive Hebb synapse" (CHS) used in
Boltzmann machine learning makes it an attractive model for real biological
synapses. The slow learning exhibited by the stochastic Boltzmann machine can
be greatly improved by using a mean field approximation and it has been shown
(Hinton, 1989) that the CHS also performs steepest descent in these
deterministic mean field networks. A major weakness of the learning
procedure, from a biological perspective, is that the derivation assumes
detailed symmetry of the connectivity. Using networks with purely asymmetric
connectivity, we show that the CHS still works in practice provided the
connectivity is grossly symmetrical so that if unit i sends a connection to
unit j, there are numerous indirect feedback paths from j to i. So long as the
network settles to a stable state, we show that the CHS approximates steepest
descent and that the proportional error in the approximation can be expected
to scale as 1/sqrt(N), where N is the number of connections.
------------------------------
Subject: two more Tech Reports available
From: Geoffrey Hinton <hinton@ai.toronto.edu>
Date: Fri, 29 Dec 89 14:19:15 -0500
Please do not reply to this message. If you want copies of either of
the TR's below, please send email with the TR number and your real
mail address to carol@ai.toronto.edu
________________________________________________________________________
USING SPATIAL COHERENCE AS AN INTERNAL TEACHER FOR A NEURAL NETWORK
Suzanna Becker and Geoffrey E. Hinton
Department of Computer Science
University of Toronto
Toronto, Ontario, M5S 1A4. CANADA
CRG-TR-89-7
Supervised learning procedures for neural networks have recently met
with considerable success in learning difficult mappings. So far,
however, they have been limited by their poor scaling behaviour,
particularly for networks with many hidden layers. A promising
alternative is to develop unsupervised learning algorithms by
defining objective functions that characterize the quality of an
internal representation without requiring knowledge of the desired
outputs of the system. Our major goal is to build self-organizing
network modules which capture important regularities in the
environment in a simple form. A layered hierarchy of such modules
should be able to learn in a time roughly linear in the number of
layers. We propose that a good objective for perceptual learning is
to extract higher-order features that exhibit simple coherence across
time or space. This can be done by transforming the input
representation into an underlying representation in which the mutual
information between adjacent patches of the input can be expressed in
a simple way. We have applied this basic idea to develop several
learning algorithms for discovering spatially coherent features in
images. Our simulations show that a network can discover depth of
surfaces when trained on binary random-dot stereograms with discrete
global shifts, as well as on real-valued stereograms of surfaces with
continuously varying disparities. Once a module of depth-tuned units
has developed, we show that units in a higher layer can discover a
simple form of surface interpolation of curved surfaces by learning
to predict the local depth measurement of one image patch from a
linear combination of depth measurements in nearby patches.
______________________________________________________________________
AN UNSUPERVISED INTERPRETATION OF ADAPTIVE EQUALIZATION
Steven J. Nowlan and Geoffrey E. Hinton
Department of Computer Science
University Of Toronto
CRG-TR-89-8
We present a novel interpretation of a simple, but powerful,
decision-directed adaptation algorithm used for equalization in
digital communication channels. The adaptation procedure is viewed as
a way of modifying the linear transformation from the vector input to
the scalar output of the equalizer so as to maximize the likelihood
that the outputs were produced by a generative model which consists
of two gaussians with fixed means. If we use a "hard" competition in
which each equalizer output is assumed to be generated by the closest
gaussian, modifying the linear transformation so as to minimize the
average variance of the two gaussians is equivalent to the familiar
decision-directed adaptation algorithm first proposed by Lucky
(1965). If we use a "soft" competition, in which the two gaussians
share responsibility for an equalizer output in proportion to the
probability densities of that output under each gaussian, the same
criterion of minimizing average variance leads to a new adaptation
algorithm which offers improved transient performance in distorted
channels.
------------------------------
Subject: Behavioral Neuroscience Faculty Position at Boulder
From: Paul Smolensky <pauls@neuron.Colorado.EDU>
Date: Fri, 29 Dec 89 17:06:55 -0700
Below is the job description for a faculty position in Behavioral
Neuroscience at the University of Colorado at Boulder. Our campus has
an active, collaborative, multi-disciplinary connectionist community.
Anyone interested in more information is welcome to contact us; if
you apply for the job, let us know so we can follow up.
-- Paul Smolensky & Mike Mozer
BEHAVIORAL NEUROSCIENCE POSITION
University of Colorado, Boulder
The Department of Psychology at the University of Colorado at
Boulder invites applications for a faculty position in Behavioral
Neuroscience, starting September 1990. Outstanding applicants at
any rank are encouraged to apply. This position carries with it
attractive research space and significant start-up funds. Appli-
cants should send a vita, 3 letters of recommendation, and a
statement of teaching and research interest to: Jerry Rudy,
Chairperson, Behavioral Neuroscience Search Committee, Department
of Psychology, Box 345, University of Colorado, Boulder, CO
80309. Application deadline is January 15, 1990.
------------------------------
Subject: SAB90 Call for Papers
From: Stewart Wilson <wilson@Think.COM>
Date: Tue, 02 Jan 90 14:22:12 -0500
Call for Papers
SIMULATION OF ADAPTIVE BEHAVIOR: FROM ANIMALS TO ANIMATS
An International Conference to be held in Paris
September 24-28, 1990
The object of the conference is to bring together researchers in
ethology, ecology, cybernetics, artificial intelligence, robotics,
and related fields so as to further our understanding of the behaviors
and underlying mechanisms that allow animals and, potentially,
robots to adapt and survive in uncertain environments.
The conference will focus particularly on simulation models in
order to help characterize and compare various organizational principles
or architectures capable of inducing adaptive behavior in real or
artificial animals.
Contact among scientists from diverse disciplines should contribute
to better appreciation of each other's approaches and vocabularies,
to cross-fertilization of fundamental and applied research, and
to defining objectives, constraints, and challenges for future work.
Contributions treating any of the following topics from the
perspective of adaptive behavior will receive special emphasis.
Individual and collective behaviors Autonomous robots
Action selection and behavioral Hierarchical and parallel organizations
sequences Self organization of behavioral
Conditioning, learning and induction modules
Neural correlates of behavior Problem solving and planning
Perception and motor control Goal directed behavior
Motivation and emotion Neural networks and classifier
Behavioral ontogeny systems
Cognitive maps and internal Emergent structures and behaviors
world models
Authors are requested to send two copies (hard copy only) of a
full paper to each of the Conference chairmen:
Jean-Arcady MEYER Stewart WILSON
Groupe de Bioinformatique The Rowland Institute for Science
URA686.Ecole Normale Superieure 100 Cambridge Parkway
46 rue d'Ulm Cambridge, MA 02142
75230 Paris Cedex 05 USA
France
e-mail: meyer%FRULM63.bitnet@ e-mail: wilson@think.com
cunyvm.cuny.edu
A brief preliminary letter to one chairman indicating the intention to
participate--with the tentative title of the intended paper and a list
of the topics addressed--would be appreciated for planning purposes.
For conference information, please also contact one of the chairmen.
Conference committee:
Conference Chair J.A. Meyer, S. Wilson
Organizing Committee Groupe de BioInformatique.ENS.France.
and local arrangements A. Guillot, J.A. Meyer, P. Tarroux,
P. Vincens
Program Committee L. Booker, USA R. Brooks, USA
P. Colgan, Canada P. Greussay, France
D. McFarland, UK L. Steels, Belgium
R. Sutton, USA F. Toates, UK
D. Waltz, USA
Official Language: English
Important Dates
31 May 90 Submissions must be received by the chairmen
30 June 90 Notification of acceptance or rejection
31 August 90 Camera ready revised versions due
24-28 September 90 Conference dates
------------------------------
End of Neuron Digest [Volume 6 Issue 5]
***************************************
