new books in MIT Neural Network/Connectionsm series
Jeff Elman
elman at crl.ucsd.edu
Mon May 3 23:23:25 EDT 1993
The following books have now appeared as part of the Neural Network
Modeling and Connection Series, and may be of interest to readers of
the connectionists mailing group. Detailed descriptions of each book,
along with table of contents, follow.
Jeff Elman
============================================================
Neural Network Modeling and Connectionism Series
Jeffrey Elman, editor. MIT Press/Bradford Books.
* Miikkulainen, R. "Subsymbolic Natural Language Processing
An Integrated Model of Scripts, Lexicon, and Memory"
* Mitchell, M. "Analogy-Making as Perception A Computer Model"
* Cleeremans, A. "Mechanisms of Implicit Learning Connectionist Models
of Sequence Processing"
* Sereno, M.E. "Neural Computation of Pattern Motion Modeling Stages of
Motion Analysis in the Primate Visual Cortex"
* Miller, W.T., Sutton, R.S., & Werbos, P.J. (Eds.), "Neural Networks for
Control"
* Hanson, S.J., & Olson, C.R. (Eds.) "Connectionist Modeling and Brain
Function The Developing Interface"
* Judd, S.J. "Neural Network Design and the Complexity of Learning"
* Mozer, M.C. "The Perception of Multiple Objects A Connectionist
Approach"
------------------------------------------------------------
New
Subsymbolic Natural Language Processing
An Integrated Model of Scripts, Lexicon, and Memory
Risto Miikkulainen
Aiming to bridge the gap between low-level connectionist models and
high-level symbolic artificial intelligence, Miikkulainen describes
DISCERN, a complete natural language processing system implemented
entirely at the subsymbolic level. In DISCERN, distributed neural
network models of parsing, generating, reasoning, lexical processing,
and episodic memory are integrated into a single system that learns to
read, paraphrase, and answer questions about stereotypical narratives.
Using the DISCERN system as an example, Miikkulainen introduces a
general approach to building high-level cognitive models from
distributed neural networks, and shows how the special properties of
such networks are useful in modeling human performance. In this approach
connectionist networks are not only plausible models of isolated
cognitive phenomena, but also sufficient constituents for complete
artificial intelligence systems.
Risto Miikkulainen is an Assistant Professor in the Department of
Computer Sciences at the University of Texas, Austin.
Contents: I.Overview. Introduction. Background. Overview of DISCERN. II.
Processing Mechanisms. Backpropagation Networks. Developing
Representations in FGREP Modules Building from FGREP Modules. III.
Memory Mechanisms. Self-Organizing Feature Maps. Episodic Memory
Organization: Hierarchical Feature Maps. Episodic Memory Storage and
Retrieval: Trace Feature Maps. Lexicon. IV. Evaluation. Behavior of the
Complete Model. Discussion. Comparison to Related Work. Extensions and
Future Work. Conclusions. Appendixes: A Story Data. Implementation
Details. Instructions for Obtaining the DISCERN Software.
A Bradford Book
May 1993 - 408 pp. - 129 illus. - $45.00
0-262-13290-7 MIISH
------------------------------------------------------------
New
Analogy-Making as Perception
A Computer Model
Melanie Mitchell
Analogy-Making as Perception is based on the premise that analogy-making
is fundamentally a high-level perceptual process in which the
interaction of perception and concepts gives rise to "conceptual
slippages" which allow analogies to be made. It describes Copycat,
developed by the author with Douglas Hofstadter, that models the
complex, subconscious interaction between perception and concepts that
underlies the creation of analogies.
In Copycat, both concepts and high-level perception are emergent
phenomena, arising from large numbers of low-level, parallel,
non-deterministic activities. In the spectrum of cognitive modeling
approaches, Copycat occupies a unique intermediate position between
symbolic systems and connectionist systems - a position that is at
present the most useful one for understanding the fluidity of concepts
and high-level perception.
On one level the work described here is about analogy-making, but on
another level it is about cognition in general. It explores such issues
as the nature of concepts and perception and the emergence of highly
flexible concepts from a lower-level "subcognitive" substrate.
Melanie Mitchell, Assistant Professor in the Department of Electrical
Engineering and Computer Science at the University of Michigan, is a
Fellow of the Michigan Society of Fellows. She is also Director of the
Adaptive Computation Program at the Santa Fe Institute.
Contents: Introduction. High-Level Perception, Conceptual Slippage, and
Analogy-Making in a Microworld. The Architecture of Copycat. Copycat's
Performance on the Five Target Problems. Copycat's Performance on
Variants of the Five Target Problems. Summary of the Comparisons between
Copycat and Human Subjects. Some Shortcomings of the Model. Results of
Selected "Lesions" of Copycat. Comparisons with Related Work.
Contributions of This Research. Afterword by Douglas R. Hofstadter.
Appendixes. A Sampler of Letter-String Analogy Problems Beyond Copycat's
Current Capabilities. Parameters and Formulas. More Detailed
Descriptions of Codelet Types.
A Bradford Book
May 1993 - 382 pp. - 168 illus. - $45.00
0-262-13289-3 MITAH
------------------------------------------------------------
New
Mechanisms of Implicit Learning
Connectionist Models of Sequence Processing
Axel Cleeremans
What do people learn when they do not know that they are learning? Until
recently all of the work in the area of implicit learning focused on
empirical questions and methods. In this book, Axel Cleeremans explores
unintentional learning from an information-processing perspective. He
introduces a theoretical framework that unifies existing data and models
on implicit learning, along with a detailed computational model of human
performance in sequence-learning situations.
The model, based on a simple recurrent network (SRN), is able to predict
the successive elements of sequences generated from finite-state
grammars. Human subjects are shown to exhibit a similar sensitivity to
the temporal structure in a series of choice reaction time experiments
of increasing complexity; yet their explicit knowledge of the sequence
remains limited. Simulation experiments indicate that the SRN model is
able to account for these data in great detail. Other architectures
that process sequential material are considered. These are contrasted
with the SRN model, which they sometimes outperform. Considered
together, the models show how complex knowledge may emerge through the
operation of elementary mechanisms - a key aspect of implicit learning
performance.
Axel Cleeremans is a Senior Research Assistant at the National Fund for
Scientific Research, Belgium.
Contents: Implicit Learning: Explorations in Basic Cognition. The SRN
Model: Computational Aspects of Sequence Processing. Sequence Learning
as a Paradigm for Studying Implicit Learning. Sequence Learning: Further
Explorations. Encoding Remote Control. Explicit Sequence Learning.
General Discussion.
A Bradford Book
April 1993 - 227 pp. - 60 illus. - $30.00
0-262-03205-8 CLEMH
------------------------------------------------------------
New
Neural Computation of Pattern Motion
Modeling Stages of Motion Analysis in the Primate Visual Cortex
Margaret Euphrasia Sereno
How does the visual system compute the global motion of an object from
local views of its contours? Although this important problem in
computational vision (also called the aperture problem) is key to
understanding how biological systems work, there has been surprisingly
little neurobiologically plausible work done on it. This book describes
a neurally based model, implemented as a connectionist network, of how
the aperture problem is solved. It provides a structural account of the
model's performance on a number of tasks and demonstrates that the
details of implementation influence the nature of the computation as
well as predict perceptual effects that are unique to the model. The
basic approach described can be extended to a number of different
sensory computations.
"This is an important book, discussing a significant and very general
problem in sensory processing. The model presented is simple, and it is
elegant in that we can see, intuitively, exactly why and how it works.
Simplicity, clarity and elegance are virtues in any field, but not often
found in work in neural networks and sensory processing. The model
described in Sereno's book is an exception. This book will have a
sizeable impact on the field." - James Anderson, Professor, Department
of Cognitive and Linguistic Sciences, Brown University
Contents: Introduction. Computational, Psychophysical, and
Neurobiological Approaches to Motion Measurement. The Model. Simulation
Results. Psychophysical Demonstrations. Summary and Conclusions.
Appendix: Aperture Problem Linearity.
A Bradford Book
March 1993 - 181 pp.- 41 illus. - $24.95
0-262-19329-9 SERNH
------------------------------------------------------------
Neural Networks for Control
edited by W. Thomas Miller, III, Richard S. Sutton,
and Paul J. Werbos
This book brings together examples of all of the most important
paradigms in artificial neural networks (ANNs) for control, including
evaluations of possible applications. An appendix provides complete
descriptions of seven benchmark control problems for those who wish to
explore new ideas for building automatic controllers.
Contents: I.General Principles. Connectionist Learning for Control: An
Overview, Andrew G. Barto. Overview of Designs and Capabilities, Paul J.
Werbos. A Menu of Designs for Reinforcement Learning Over Time, Paul J.
Werbos. Adaptive State Representation and Estimation Using Recurrent
Connectionist Networks, Ronald J. Williams. Adaptive Control using
Neural Networks, Kumpati S. Narendra. A Summary Comparison of CMAC
Neural Network and Traditional Adaptive Control Systems, L. Gordon
Kraft, III, and David P. Campagna. Recent Advances in Numerical
Techniques for Large Scale Optimization, David F. Shanno. First Results
with Dyna, An Integrated Architecture for Learning, Planning and
Reacting, Richard S. Sutton.
II. Motion Control. Computational Schemes and Neural Network Models for
Formation and Control of Multijoint Arm Trajectory, Mitsuo Kawato.
Vision-Based Robot Motion Planning, Bartlett W. Mel. Using Associative
Content-Addressable Memories to Control Robots, Christopher G. Atkeson
and David J. Reinkensmeyer. The Truck Backer-Upper: An Example of
Self-Learning in Neural Networks, Derrick Nguyen and Bernard Widrow. An
Adaptive Sensorimotor Network Inspired by the Anatomy and Physiology of
the Cerebellum, James C. Houk, Satinder P. Singh, Charles Fisher, and
Andrew G. Barto. Some New Directions for Adaptive Control Theory in
Robotics, Judy A. Franklin and Oliver G. Selfridge.
III. Application Domains. Applications of Neural Networks in Robotics
and Automation for Manufacturing, Arthur C. Sanderson. A Bioreactor
Benchmark for Adapive Network-based Process Control, Lyle H. Ungar. A
Neural Network Baseline Problem for Control of Aircraft Flare and
Touchdown, Charles C. Jorgensen and C. Schley. Intelligent Conrol for
Multiple Autonomous Undersea Vehicles, Martin Herman, James S. Albus,
and Tsai-Hong Hong. A Challenging Set of Control Problems, Charles W.
Anderson and W. Thomas Miller.
A Bradford Book
1990 - 524 pp. - $52.50
0-262-13261-3 MILNH
------------------------------------------------------------
Connectionist Modeling and Brain Function
The Developing Interface
edited by Stephen Jose Hanson and Carl R. Olson
This tutorial on current research activity in connectionist-inspired
biology-based modeling describes specific experimental approaches and
also confronts general issues related to learning, associative memory,
and sensorimotor development.
"This volume makes a convincing case that data-rich brain scientists and
model-rich cognitive psychologists can and should talk to one another.
The topics they discuss together here - memory and perception - are of
vital interest to both, and their collaboration promises continued
excitement along this new scientific frontier." - George Miller,
Princeton University
Contents: Part I: Overview. Introduction: Connectionism and
Neuroscience, S. J. Hanson and C. R. Olson. Computational Neuroscience,
T. J. Sejnowski, C. Koch, and P. S. Churchland. Part II: Associative
Memory and Conditioning. The Behavioral Analysis of Associative Learning
in the Terrestrial Mollusc Limax Maximus: The Importance of Inter-event
Relationships, C. L. Sahley. Neural Models of Classical Conditioning: A
Theoretical Viewpoint, G. Tesauro. Unsupervised Perceptual Learning: A
Paleocortical Model, R. Granger, J. Ambros-Ingerson, P. Anton, and G.
Lynch. Part III. The Somatosensory System. Biological Constraints on a
Dynamic Network: The Somatosensory Nervous System, T. Allard. A Model of
Receptive Field Plasticity and Topographic Reorganization in the
Somatosensory Cortex, L. H. Finkel. Spatial Representation of the Body,
C. R. Olson and S. J. Hanson. Part IV: The Visual System. The
Development of Ocular Dominance Columns: Mechanisms and Models. K. D.
Miller and M. P. Stryker. Self- Organization in a Perceptual System: How
Network Models and Information Theory May Shed Light on Neural
Organization, R. Linsker. Solving the Brightness-From-Luminance Problem:
A Neural Architecture for Invariant Brightness Perception, S. Grossberg
and D. Todorovic.
A Bradford Book
1990 - 423 pp. - $44.00
0-262-08193-8 HANCH
------------------------------------------------------------
Neural Network Design and the Complexity of Learning
J. Stephen Judd
Using the tools of complexity theory, Stephen Judd develops a formal
description of associative learning in connectionist networks. He
rigorously exposes the computational difficulties in training neural
networks and explores how certain design principles will or will not
make the problems easier.
"Judd . . . formalized the loading problem and proved it to be
NP-complete. This formal work is clearly explained in his book in such a
way that it will be accessible both to the expert and nonexpert." - Eric
B. Baum, IEEE Transactions on Neural Networks
"Although this book is the true successor to Minsky and Papert's
maligned masterpiece of 1969 (Perceptrons), Judd is not trying to
demolish the field of neurocomputing. His purpose is to clarify the
limitations of a wide class of network models and thereby suggest
guidelines for practical applications." - Richard Forsyth, Artificial
Intelligence & Behavioral Simulation
Contents: Neural Netowrks: Hopes, Problems, and Goals. The Loading
Problem. Other Studies of Learning. The Intractability of Loading.
Subcases. Shallow Architectures. Memorization and Generalization.
Conclusions. Appendices
A Bradford Book
1990 - 150 pp. - $27.50
0-262-10045-2 JUDNH
------------------------------------------------------------
The Perception of Multiple Objects
A Connectionist Approach
Michael C. Mozer
Building on the vision studies of David Marr and the connectionist
modeling of the PDP group it describes a neurally inspired computational
model of two-dimensional object recognition and spatial attention that
can explain many characteristics of human visual perception. The model,
called MORSEL, can actually recognize several two-dimensional objects at
once (previous models have tended to blur multiple objects into one or
to overload). Mozer's is a fully mechanistic account, not just a
functional-level theory.
"Mozer's work makes a major contribution to the study of visual
information processing. He has developed a very creative and
sophisticated new approach to the problem of visual object recognition.
The combination of computational rigor with thorough and knowledgeable
examination of psychological results is impressive and unique." - Harold
Pashler, University of California at San Diego
Contents: Introduction. Multiple Word Recognition. The Pull-Out Network.
The Attentional Mechanism. The Visual Short-Term Memory. Psychological
Phenomena Explained by MORSEL. Evaluation of MORSEL. Appendixes: A
Comparison of Hardware Requirements. Letter Cluster Frequency and
Discriminability Within BLIRNET's Training Set.
A Bradford Book
1991 - 217 pp - $27.50
0-262-13270-2 MOZPH
-------------------------------------------------------------
ORDER FORM
Please send me the following book(s):
Qty Author Bookcode Price
___ Cleeremans CLEMH 30.00
___ Hanson HANCH 44.00
___ Judd JUDNH 27.50
___ Mikkulainen MIISH 45.00
___ Miller MILNH 52.50
___ Mitchell MITAH 45.00
___ Mozer MOZPH 27.50
___ Sereno SERNH 24.95
___ Payment Enclosed ___ Purchase Order Attached
Charge to my ___ Master Card ___ Visa
Card# _______________________________
Exp.Date _______________
Signature _________________________________________________
_____ Total for book
$2.75 Postage
_____ Please add 50c postage for each additional book
_____ Canadian customers Add 7% GST
_____ TOTAL due MIT Press
Send To:
Name ______________________________________________________
Address ___________________________________________________
City ________________________ State ________ Zip __________
Daytime Phone ________________ Fax ________________________
Make checks payable and send order to:
The MIT Press * 55 Hayward Street * Cambridge, MA 02142
For fastest service call (617) 625-8569
or toll-free 1-800-356-0343
The MIT Guarantee: If for any reason you are not completely satisfied,
return your book(s) within ten days of receipt for a full refund or
credit.
3ENET
More information about the Connectionists
mailing list