Copy Link
Add to Bookmark
Report
Neuron Digest Volume 11 Number 28
Neuron Digest Monday, 26 Apr 1993 Volume 11 : Issue 28
Today's Topics:
New Books from Academic Press
"Brain Usage"
Send submissions, questions, address maintenance, and requests for old
issues to "neuron-request@cattell.psych.upenn.edu". The ftp archives are
available from cattell.psych.upenn.edu (130.91.68.31). Back issues
requested by mail will eventually be sent, but may take a while.
----------------------------------------------------------------------
Subject: new books
From: Kathleen Tibbetts <ktibbetts@igc.apc.org>
Date: Thu, 15 Apr 93 12:37:28 -0800
[[ Editor's Note: As faithful readers know, this Digest is not generally
an appropriate place for commercial announcements. However, simple
descriptions without marketing hyperbole seems reasonable to include,
especially (in my opinion) books and other publications. -PM ]]
Following are announcements for two new books from AP that
may be of interest to your readers. Please let me know if I can
provide any additional information.
Sincerely,
Kathleen M. Tibbetts
Acquisitions Editor
Computer Science
- ------------------------------------------------------------
Academic Press Phone: (617) 876-3901 x107
955 Massachusetts Avenue Fax: (617) 661-3608
Cambridge, MA 02139 USA email: ktibbetts@igc.org
- ------------------------------------------------------------
Academic Press announces the publication of:
PRACTICAL NEURAL NETWORK RECIPES IN C++
by
Timothy Masters, Ph.D.
Designed for those with no previous knowlegde of neural networks,
PRACTICAL NEURAL NETWORK RECIPES IN C++ serves as a cookbook for neural
network solutions to practical problems. It will enable those with
moderate programming experience to select a neural network model
appropriate to solving a particular problem and to produce a working C++
program. The book provides guidance along the entire problem-solving
path, including designing the training set, preprocessing variables,
training and validating the network, and evaluating its performance. A
high-density IBM diskette bound in the book includes the source code for
all programs in the book.
Key Features:
* Provides a lengthy treatment of practical fuzzy logic and examples of
hybrid fuzzy/neural models. Complete code for implementing all major fuzzy
operations is shown.
* Includes algorithms for implementing two popular stochastic optimization
techniques -- simulated annealing and genetic optimization.
* Includes a detailed disciussion of computation of decisions confidences.
* Covers feature identification in detail
Contents:
1. Foundations
2. Classification
3. Autoassociation
4. Time Series Prediction
5. Function Approximation
6. Multilayer Feedforward Networks
7. Eluding Local Minima I: Simulated Annealing
8. Eluding Local Minima II: Genetic Optimization
9. Regression and Neural Networks
10. Designing Feedforward Network Architectures
11. Interpreting Weights: How Does This Thing Work?
12. Probablistic Neural Networks
13. Functional Link Networks
14. Hybrid Networks
15. Designing the Training Set
16. Fuzzy Data and Processing
17. Unsupervised Training
18. Evaluating Perfomance of Neural Networks
19. Confidence Measures,
20. Optimizing The Decision Threshold
21. Using the NEURAL Program
Appendix: Source code listings
Bibliography
Index
ISBN: 0-12-479040-2 $44.95 paperback March 1993 493 pp.
U.S. and Canadian customers may call toll-free 1-800-321-5068 or
fax 1-800-336-7377 Mon. -Fri. 8:30 AM to 7:00 PM Esatern Time.
Free shipping and handling with prepaid orders.
Visa, Mastercard, and American Express accepted, or send check or
money order to:
Academic Press
HB Order Fulfillment Department #18182
6277 Sea Harbor Drive
Orlando, FL 32887
In Europe call: 081-300-3322
Or write:
Academic Press
Book Marketing Department
24-48 Oval Road
London NW1 7DX, U.K.
Academic Press announces the publication of:
BIOLOGICAL NEURAL NETWORKS IN
INVERTEBRATE NEUROETHOLOGY AND ROBOTICS
Edited by
Randall D. Beer and Roy Ritzmann
Case Western Reserve University, Cleveland, Ohio
Thomas McKenna
Biological Intelligence Program, Office of Naval Research,
Arlington, Virginia
A Volume in the NEURAL NETWORKS: FOUNDATIONS TO APPLICATIONS series
This is the first book to integrate research by neuroethologists
interested in the neural basis of natural animal behavior and roboticists
interested in building versatile and robust robots. Biological Neural
Networks in Invertebrate Neuroethology and Robotics contains 17 essays
that survey neural control of movement and orientation, describe computer
models and neural control circuits, and give examples of actual robot
implementations. This book is the second volume in Academic Press' new
series Neural Networks: Foundations to Applications, which seeks to
emphasize the interdisciplinary exchange of ideas that is central to
advances in neural networks research.
Key Features:
* Presents for the first time the results of research at the intersection
of the fields of neuroethology and robotics
* Emphasizes potential advances for both biologists and engineers
CONTENTS
I. Neuroethology Control of Leg Movement
Integration of Individual Leg Dynamics with Whole Body Movement
in Anthropod Locomotion
R.J. Full
The Walking of Cockroaches--Deceptive Simplicity
F. Delcomyn
Load Compensatory Reactions in Insects: Swaying and Stepping
Strategies in Posture and Locomotion
S.N. Zill
Integration by Spiking and Nonspiking Local Neuron in the Locust
Central Nervous System: The Importance of Cellular and Synaptic
Properties for Network Function
G. Laurent
II Neuroethology II: Control of Orientation
Multisensory Processing of Movement: Antennal and Cercal
Mediation of Escape Turning in the Cockroach
C.M. Comer and J.P. Dowd
The Neural Organization of the Cockroach Escape and Its Role in
Context Dependent Orientation
R.E. Ritzmann
Acoustic Startle: An Adaptive Behavioral Act in Flying Insects
R.R. Roy
Organization of Goal-Oriented Locomotion: Pheromone-Modulated
Flight Behavior of Moths
E.A. Arbas, M.A. Willis, and R. Kanzaki
A New Role for the Insect Mushroom Bodies: Place Memory and
Motor Control
N. Strausfeld, M. Mizunami and J. M. Weibrecht
III Computer Modeling
Modeling a Reprogrammable Central Pattern Generating Network
A.I. Selverston, P. Rowat and M.E.T. Boyle
Voyages Through Weight Space: Network Models of an Escape Reflex
in the Leech
S.R. Lockery and T.J. Sejnowski
Simulations of Cockroach Locomotion and Escape
R.D. Beer and H.I. Chiel
Lobster Walking As a Model for an Omnidirectional Robotic
Ambulation Architecture
J. Ayers and J. Crisman
IV Robotics
Legged Robots
M.H. Raibert and J.K. Hodgins
A Robot that Walks: Emergent Behavior from a Carefully Evolved
Network
R. Brooks
Control of a Hexapod Robot Using a Biologically Inspired Neural
Network
R.D. Quinn and K.S. Espenschied
Modeling Neural Function at the Scheme Level: Implications and
Results of the Robotic Control
R.C. Arkin
ISBN 0-12-084728-0 $64.95 hardcover October, 1992 400 pp.
U.S. and Canadian Customers may call toll-free 1-800-321-5068
or Fax 1-800-336-7377 Mon. - Fri. 8:30 Am to 7:00 PM Eastern Time.
Free shipping and handling with prepaid orders.
Visa, Mastercard, and American Express accepted, or send
check or money order to:
Academic Press
HB Order Fulfillment Department #18182
6277 Sea Harbor Drive
Orlando, FL 32887
In Europe call: 081-300-3322
Or write:
Academic Press
Book Marketing Department
24-48 Oval Road
London NW1 7DX, U.K.
------------------------------
Subject: "Brain Usage"
From: MERKLEY@delphi.com
Date: 08 Apr 93 21:05:19 -0500
[[ Editor's Note: Thanks for this interesting compendium. While a bit
afield from our regular topics, it does bring up the question of ANN (and
natural neural net) capacity -- a topic occaisionally discussed but
rarely in depth. I hope responses might also be directed back to this
Digest (in addition to the BARIN-L list). -PM ]]
In the category of 'half-baked ideas' :-), the following discussion has
taken on a life of its own on BRAIN-L. Any takers? Please respond, if
you can, to BRAIN-L@VM1.MCGILL.CA
From: IN%"NICE%BUTLERU.BITNET@VM1.MCGILL.CA" "Brian C. Nice" 2-APR-1993 13:38:49.51
Subj: Brain usage
Does anyone have any numbers on what percentage of our brain's full
potential that we as humans use? Any documentation would also be greatly
appreciated. I am not on this list, so please respond to me directly.
Thanks in advance!
Brian
From: IN%"Langdon@GANDLF.UINDY.EDU" "John Langdon" 2-APR-1993 15:49:27.09
Subj: RE: Brain usage
In message "Brian C. Nice" writes:
> Does anyone have any numbers on what percentage of our brain's full potential
> that we as humans use? Any documentation would also be greatly appreciated.
> I am not on this list, so please respond to me directly. Thanks in advance!
I have many students who are using only minute portions of their brains'
full potentials, but until someone has a method of measuring/defining
this concept, one would have a difficult time quantifying it. I have seen
this argument expressed as a percentage of neurons used. That would be a
little more concrete, but it only is the inverse of the question "To what
percentage of the brain cells can we not assign a known function?"
In other words, I don't believe the question you cite has any meaning.
**********************************
John H. Langdon email LANGDON@GANDLF.UINDY.EDU
Department of Biology office phone (317)788-3447
University of Indianapolis FAX (317)788-3569
1400 East Hanna Avenue
Indianapolis, IN 46227
From: IN%"moriem@PIKE.EE.MCGILL.CA" "Morie Malowany" 2-APR-1993 17:46:17.13
Subj: RE: Brain usage
The question of what percentage of the brain's computational power is
being utilized at time t, or what it is capable of in some limit or
asymptotic sense has no meaning ...
Yet. It is an interesting question though, how to quantify the
computational power of a biological system in a meaningful way (other
that just the information theoretic sense of how many possible states
does it have, if each neuron can be a zero or a one ...)
- -Morie.
Montreal, Quebec
Canada
From: IN%"X042%HECMTL01.BITNET@VM1.MCGILL.CA" 5-APR-1993 08:29:27.05
Subj: RE: Brain usage
I seem to recall that Einstein did say something about this. But don't
forget he was a physician ! I guess we cannot answer to this question
(and certainly not be able for a long, long time). And by the way, what
really is brain's full potential ?
+--------------------------------------------------------------------------+
| JACQUES BRISSON Ecole des Hautes Etudes Commerciales (HEC) |
| X042@HEC.CA (internet) 5255 Decelles (QM3333-S610) |
| X042@HECMTL01.BITNET(bitnet) Montreal, Quebec, CANADA, H3T 1V6 |
+--------------------------------------------------+-----------------------+
| ...for the first time, it seems there may be hope of putting the slice |
| back in the brain, not by the art of transplantation, but by the art of |
| of computation. |
| B.L. Mc Naughton |
+--------------------------------------------------------------------------+
From: IN%"SCOTT@HEP.PHYSICS.MCGILL.CA" 5-APR-1993 11:28:13.91
Subj: RE: Brain usage
There seems to be a maddening amount of popular attention devoted to this
question of the percentage of our brain potential that is utilized. The
question is premature at best, if not entirely ill-posed. I feel
confident that no one in any field related to the brain is likely to
claim quantative knowledge of how the brain functions, and I expect that
there would be unanimous concensus that its "potential" is not simply
related to the number of neurons (beyond that this should exceed some
minimum) or the percentage active during a particular interval. Moreover
the question of what it is the brain does seems to have remained
impervious to all attempts to denotatively categorize it. The concept of
"intelligence" for instance is notoriously slippery. Given that we don't
know what it does or how it does it, a yardstick for the brain seems
mighty remote
Scott
Agree with all previous responders who have stated, in various ways, that
the brain's potential is unlikely to be a simple function of the number
of neurons it has or how many of them are individually active.
But to the extent that what it is that the brain does,
neurophysiologically, is nowadays thought to be parallel distributed
processing, we may at any rate be a step closer to asking the right sort
of question about the brain's capacity if we take the connectionist
position and ask something like "how many discrete points in its
particular state space is this neural network capable of discriminating?"
Agree with all who state that this is pragmatically an impossible
question for the forseeable future; nevertheless, since those who study
simple artificial neural nets are able to answer mathematically, and with
a straight face, questions about what really determines how much
information a given system is capable of processing and storing, we might
see if there are any connectionist buffs out there willing to stick their
necks out and approach Brian's original question.
By the way, Brian said he wasn't on this list -- has anyone kept him
apprised of the responses his question has generated? Since I raised
this particular question I guess I'll drop him a line or two myself.
Matt Merkley
The Menninger Clinic
Topeka, Kansas, USA
From: IN%"cris@HEBB.PSYCH.MCGILL.CA" "Cristina Sorrentino" 6-APR-1993 13:28:49.95
Subj: brain usage
For what it's worth, the question of brain usage is not so far fetched,
ill-posed, impossible to think about until some far distant date in the
future, as some list members have suggested. For example, during his
talk on April 1st here at McGill on the evolution of language, Pinker
mentioned one victim of hydroencephalus who has above normal intelligence
(ie. is a member of the Oxford Debating club etc.) Hydroencephalus is a
condition where a person's brain is mostly ventricles and ventricle
fluid, with a normal brain stem and very, very little cortex. Typically,
victims of this disorder do not have normal intelligence. But it is not
unheard of that people with the disorder have normal or above normal
intelligence.
Why do we care? Well, this is just the type of phenomenon which can give
us a lead to the question of brain usage, suggesting that the issue is
not intractable. For one thing, it IMPLIES that much cortex is simply
redundant. It SUGGESTS that we could quantitatively if not qualitatively
look into the question of brain usage and brain function by studying the
brains of hydroenchephalics...
Comments?
Cristina Sorrentino
From: IN%"MYERS2@HEP.PHYSICS.MCGILL.CA" "James Anglin" 7-APR-1993 15:12:51.24
Subj: RE: Brain usage
By way of agreement with John Langdon's question, but wistfully, I offer
a measure of brain potential: the Godel. As a reference standard, I
consider Kurt Godel's theorems about incompleteness. I estimate that
there have been no nore than about 100 such intellectual achievements in
history. There have been about ten billion human beings. So on average
we use something like one hundred-millionth of our potential.
No wonder my supervisor thinks I'm lazy.
James.
From: IN%"YB839C@GWUVM.GWU.EDU" "John Opfer" 7-APR-1993 19:26:10.78
Subj: Potentiality and the Brain
James deduces our brain potentiality on the basis of the sum of
intellectual achievements of mankind and the number of actual people who
have lived. This simply isn't logical. The potentiality for
intellectual achievement is something that isn't common. There is no
collective brain. Dividing such a mythological organ by the number of
actual men in order to determine the potential for achievement of each
individual man is the fallacy of division, as logicians put it. This is
the same fallacy as trying to deduce the cumulative effects of M&Ms
dropping on your foot over a thousand years from one's knowledge of a
safe dropping on your head. The difference is great: it is difference of
life or death. In the case of brain potentiality, it is the difference
between Galileo and the village idiot.
John Opfer
The George Washington University
From: IN%"leslie@BINKLEY.CS.MCGILL.CA" "Leslie DAIGLE" 8-APR-1993 00:40:31.50
Subj: Re. potentiality etc etc
In a tone of impish political correctness, that seems to suit the whole
situation:
>From John Opfer's message:
> mythological organ by the number of actual men in order to determine the
> potential for achievement of each individual man is the fallacy of division,
... to say nothing of how we might measure _women's_ contributions,
potentials, and achievements... :-)
Have a nice day, eh?!
Leslie.
- ---------------------------------------------------------------------
"If you think I've given a silly answer, perhaps Leslie Daigle
you should reconsider the question you asked." leslie@cs.mcgill.ca
--ThinkingCat Montreal, Canada
- ---------------------------------------------------------------------
From: IN%"$HWG%PCCJES2.BITNET@VM1.MCGILL.CA" "Harold Gordon" 8-APR-1993 09:23:27.92
Subj: large brains...
The amount of commentary the issue of brain potential has generated is, well,
mind-boggling. Should Churchland have a chapter on this in her next book?
Harold Gordon
P.S. Perhaps the list is starved for something to talk about?
From: IN%"moriem@PIKE.EE.MCGILL.CA" "Morie Malowany" 8-APR-1993 09:33:38.92
Subj: RE: large brains...
I for one am very interested in a serious approach to this brain
potential question, even if it cannot be definitively answered at the
current level of knowledge. How might it best be approached? Suggestions
of a connectionist approach and a study of anomolous brains with their
level of functionalities, etc that have appeared on the list seem
promising. Any others? What do you think Patricia Churchland would say
about brain potential?
- -Morie
Montreal, Quebec
Canada
From: IN%"$HWG%PCCJES2.BITNET@VM1.MCGILL.CA" "Harold Gordon" 8-APR-1993 13:38:32.33
Subj: potential capacity
I have to admit that beyond the popular question of what percentage of our
brains are we using, is a very good question is "can we calculate (estimate)
a potential capacity?" I never thought about it seriously before, let me
think out loud:
Storage is likely dependent on functional connections between neurons,
and neuronal systems. So number of neurons, number (and quality) of
connections are certainly a limiting feature. Number of connections
depends on the number of branches of the dendritic tree and the number of
synaptic boutons. Similarly, development of functional synapses
(receptors, metabolic actions for the transmitters, etc.) are also
required. There must be a mathematical model for the number of
possibilities once one decides how many functional connections is
necessary for an item to be "set" in storage. I would guess that items
are fuzzily set in storage. This means that a stimulus to recall the
item would be successful all the time for "ingrained" items and some of
the time for fuzzy items. This fuzziness is probably dependent on
redundancy of the recall network which may be organized by category
(semantic--a fruit but not a pear), color, size, taste, aroma,...
How do individuals differ in their storage capacity? By number of
neurons and connections, but probably also in their category organizers
and inter- connections between them (ref. perfect pitch discussion). Not
everyone organizes their categories in the same way (biologically
speaking as well as learned efficiency), so we all have different
capacities in BOTH amount and type.
That brings to mind of how people differ in terms of thinking modes. One
whose brain is a supreme organizer may be very compartmentalized and
cannot integrate categories as easily as a less "organzied" brain. Thus,
a less organized brain may come up with more "creative" ideas but cannot
remember how to spell brain. (The observation is true, I just made up
the explanation). So capacity is also dependent on definition of what is
being counted--items remembered or ideas generated.
How's that for a start?
Harold Gordon
From: IN%"forb0004@STUDENT.TC.UMN.EDU" "Eric J. Forbis" 8-APR-1993 14:01:12.21
Subj: RE: Potentiality and the Brain
On the question of how well "we" use our brains, shouldn't we first
define what portions of the brain are responsible for the sense of
identity, self, then measure their use? Most of the brain is dedicated to
processing visual stimuli, sound, breathing, etc.; these functions may
feed into that region responsible for identity, but of themselves can't
be said to have independent teleological goals, nor can we directly use
them.
I suspect that when we winnow out all but regions dedicated to cognitive
function, we'll be impressed with what can be accomplished with so few
neurons.
Eric J. Forbis forb0004@student.tc.umn.edu
eric@mermaid.micro.umn.edu
From: IN%"BHALL@VM2.YORKU.CA" "B. Hall" 8-APR-1993 14:02:58.95
Subj: potential capacity
In regards of Harold Gordon's on brain potential, is there any validity
to the scenario in the recent movie "Lawnmower Man"?
To those of you who haven't seen the movie, a scientist used a
combination of a virtual environment (computer generated) and drugs to
stimulate his subjects brain thus increasing his intelligence
dramatically.
Now I know this is a bit farfetched but what I'm interested in
specifically is the aspect of the virtual/subliminal(?) training. Is
there any scientific basis for this?
BEST REGARDS, B.S. HALL
(YES THAT IS MY REAL NAME)
From: IN%"moriem@PIKE.EE.MCGILL.CA" "Morie Malowany" 8-APR-1993 14:11:04.73
Subj: storage capacity
H. Gordon's discussion of storage capacity reminded me of an oft-cited
paper in the neural-nets literature,
R.J. McEliece, E.C. Posner, E.R. Rodemid, S.S. Venkatesh, ``The Capacity
of the Hopfield Associative Memory,'' _IEEE Trans. Info. Theory_,
Vol. IT-33, pp. 1-33, July 1987.
which derives a result to the effect that, if I recall correctly,
... the number of well-behaved (in a dynamical
systems sense) stored memories in a binary-valued hopfield-type neural
network, having N neurons in a single-layer fully-feedback-connected
configuration, is
# memories <= 0.1N
However, the ramifications for the brain of this result are perhaps
negligible. It just points out that the relation between number of
states one can represent, and number that can be made useful as
"associative memories" is generally not very straightforward.
F.Y.I. (with my apologies to the purists for any oversimplifications)
- -Morie
Montreal, Quebec
Canada
From: IN%"moriem@PIKE.EE.MCGILL.CA" "Morie Malowany" 8-APR-1993 14:19:24.85
Subj: RE: potential capacity
Re: the movie "Lawnmower Man", perhaps this is a bit off-topic from the
virtual learning issue, but rather in the sense of making virtual reality
computing feel more real, I found the idea of providing the
computer-person link with a
"direct connection to the human nervous or endocrine system?"
rather intriguing. Of course, the potential dangers of what happens to a
user connected to the system when an error condition occurs are rather
disturbing. But I suppose it is no worse than other "critical computing"
applications, such as nuclear power plant control, or space shuttle
navigation / launch/ lifesupport
etc.
I don't suppose anyone knows of any actual research papers on such an
interface?
From: IN%"kuslikia@GVSU.EDU" "AL KUSLIKIS" 8-APR-1993 17:55:02.21
Subj: brain potential
In line with the connectionist suggestion regarding determining brain
information-crunching potential, I think that it might be interesting to
try to determine (or at least to pretend that such a determination is
possible) what human representational space encompasses. In other words,
what is the range of representational states that the uniquely human
repertoire of sensory processes and various associative mechanisms makes
possible? The average person's typical day represents a subset of that
range, I tend to think.
I'm thinking of "representation" at the bare minimum as a pattern of
activation of a group of interacting neurons. It's a re-presentation
because that pattern presumably has a relationship to another pattern, i.e.
of some sort of input, which may include former activation states of the
neuron group instantiating the representation. Representations, especially
at the level of "consciousness" involve hierarchically related patterns
described by sets of sets of interacting neurons. "Thinking" generally
describes a habituated sequence of transitions between representational
states. One could ask, how many types of thought are possible, or more
specifically, what humanly-comprehensible logical systems are possible? An
answer to that would (sort-of) give us brain potential, I'd say.
Al K.
------------------------------
End of Neuron Digest [Volume 11 Issue 28]
*****************************************
