| By Author | [ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z | Other Symbols ] |
| By Title | [ A B C D E F G H I J K L M N O P Q R S T U V W X Y Z | Other Symbols ] |
| By Language |
|
Download this book: [ ASCII ] Look for this book on Amazon Tweet |
Title: Self-Organizing Systems, 1963
Author: Various
Language: English
As this book started as an ASCII text book there are no pictures available.
*** Start of this LibraryBlog Digital Book "Self-Organizing Systems, 1963" ***
Transcriber’s Notes:
Underscores “_” before and after a word or phrase indicate _italics_
in the original text.
Equal signs “=” before and after a word or phrase indicate =bold= in
the original text.
A single underscore after a symbol indicates a subscript.
Carat symbol “^” designates a superscript.
A bold arrow in front of a letter indicates a vector,
e.g. ⮕M means “the vector M”.
Small capitals have been converted to SOLID capitals.
Illustrations have been moved so they do not break up paragraphs.
Typographical and punctuation errors have been silently corrected.
SELF-ORGANIZING
SYSTEMS
1963
=Edited By=
=JAMES EMMETT GARVEY=
_Office of Naval Research
Pasadena, California_
=ACR-96=
=OFFICE OF NAVAL RESEARCH
DEPARTMENT OF THE NAVY
WASHINGTON, D.C.=
For sale by the Superintendent of Documents.
U.S. Government Printing Office
Washington, D.C., 20402—Price $1.50
CONTENTS
Foreword iv
The Ionic Hypothesis and Neuron Models 1
—E. R. Lewis
Fields and Waves in Excitable Cellular Structures 19
—R. M. Stewart
Multi-Layer Learning Networks 37
—R. A. Stafford
Adaptive Detection of Unknown Binary Waveforms 46
—J. J. Spilker, Jr.
Conceptual Design of Self-Organizing Machines 52
—P. A. Kleyn
A Topological Foundation for Self-Organization 65
—R. I. Ścibor-Marchocki
On Functional Neuron Modeling 71
—C. E. Hendrix
Selection of Parameters for Neural Net Simulations 76
—R. K. Overton
Index of Invited Participants 77
FOREWORD
The papers appearing in this volume were presented at a Symposium
on Self-Organizing Systems, which was sponsored by the Office of
Naval Research and held at the California Institute of Technology,
Pasadena, California, on 14 November 1963. The Symposium was organized
with the aim of providing a critical forum for the presentation and
discussion of contemporary significant research efforts, with the
emphasis on relatively uncommon approaches and methods in an early
state of development. This aim and nature dictated that the Symposium
be in effect a Working Group, with numerically limited invitational
participation.
The papers which were presented and discussed did in fact serve
to introduce several relatively unknown approaches; some of the
speakers were promising young scientists, others had become known for
contributions in different fields and were as yet unrecognized for
their recent work in self-organization. In addition, the papers as a
collection provided a particularly broad, cross-disciplinary spectrum
of investigations which possessed intrinsic value as a portrayal of
the bases upon which this new discipline rests. Accordingly, it became
obvious in retrospect that the information presented and discussed at
the Symposium was of considerable interest—and should thus receive
commensurate dissemination—to a much broader group of scientists and
engineers than those who were able to participate directly in the
meeting itself. This volume is the result of that observation; as an
edited collection of the papers presented at the Symposium, it forms
the Proceedings thereof. If it provides a useful reference for present
and future investigators, as well as documenting the source of several
new approaches, it will have fulfilled its intended purpose well.
A Symposium which takes the nature of a Working Group depends for its
utility especially upon effective commentary and critical analysis,
and we commend all the participants for their contributions in this
regard. It is appropriate, further, to acknowledge the contributions
to the success of the Symposium made by the following: The California
Institute of Technology for volunteering to act as host and for
numerous supporting services; Professor Gilbert D. McCann, Director
of the Willis Booth Computing Center at the California Institute of
Technology, and the members of the technical and secretarial staffs
of the Computing Center, who assumed the responsibility of acting as
the immediate representatives of the Institute; the members of the
Program Committee, who organized and led the separate sessions—Harold
Hamilton of General Precision, Joseph Hawkins of Ford Motor Company,
Robert Stewart of Space-General, Peter Kleyn of Northrop, and Professor
McCann; members of the Technical Information Division of the Naval
Research Laboratory, who published these Proceedings; and especially
the authors of the papers, which comprised the heart of the Symposium
and subsequently formed this volume. To all of these the sponsors wish
to express their very sincere appreciation.
JAMES EMMETT GARVEY
_Office of Naval Research Branch Office
Pasadena, California_
MARGO A. SASS
_Office of Naval Research
Washington, D.C._
The Ionic Hypothesis and Neuron Models
E. R. LEWIS
_Librascope Group, General Precision, Inc.
Research and Systems Center
Glendale, California_
The measurements of Hodgkin and Huxley were aimed at
revealing the mechanism of generation and propagation
of the all-or-none spike. Their results led to the
Modern Ionic Hypothesis. Since the publication of
their papers in 1952, advanced techniques with
microelectrodes have led to the discovery of many
modes of subthreshold activity not only in the axon
but also in the somata and dendrites of neurons. This
activity includes synaptic potentials, local response
potentials, and pacemaker potentials.
We considered the question, “Can this activity also
be explained in terms of the Hodgkin-Huxley Model?”
To seek an answer, we have constructed an electronic
analog based on the ionic hypothesis and designed
around the data of Hodgkin and Huxley. Synaptic
inputs were simulated by simple first-order or
second-order networks connected directly to simulated
conductances (potassium or sodium). The analog has,
with slight parameter adjustments, produced all modes
of threshold and subthreshold activity.
INTRODUCTION
In recent years physiologists have become quite adept at probing
into neurons with intracellular microelectrodes. They are now able,
in fact, to measure (a) the voltage change across the postsynaptic
membrane elicited by a single presynaptic impulse (see, for examples,
references 1 and 2) and (b) the voltage-current characteristics
across a localized region of the nerve cell membrane (3), (4), (5),
(6). With microelectrodes, physiologists have been able to examine
not only the all-or-none spike generating and propagating properties
of axons but also the electrical properties of somatic and dendritic
structures in individual neurons. The resulting observations have
led many physiologists to believe that the individual nerve cell
is a potentially complex information-processing system far removed
from the simple two-state device envisioned by many early modelers.
This new concept of the neuron is well summarized by Bullock in his
1959 _Science_ article (10). In the light of recent physiological
literature, one cannot justifiably omit the diverse forms of somatic
and dendritic behavior when assessing the information-processing
capabilities of single neurons. This is true regardless of the means of
assessment—whether one uses mathematical idealizations, electrochemical
models, or electronic analogs. We have been interested specifically in
electronic analogs of the neuron; and in view of the widely diversified
behavior which we must simulate, our first goal has been to find a
unifying concept about which to design our analogs. We believe we have
found such a concept in the Modern Ionic Hypothesis, and in this paper
we will discuss an electronic analog of the neuron which was based on
this hypothesis and which simulated not only the properties of the
axon but also the various subthreshold properties of the somata and
dendrites of neurons.
We begin with a brief summary of the various types of subthreshold
activity which have been observed in the somatic and dendritic
structures of neurons. This is followed by a brief discussion of the
Hodgkin-Huxley data and of the Modern Ionic Hypothesis. An electronic
analog based on the Hodgkin-Huxley data is then introduced, and we show
how this analog can be used to provide all of the various types of
somatic and dendritic activity.
SUBTHRESHOLD ELECTRICAL ACTIVITY IN NEURONS
In studying the recent literature in neurophysiology, one is
immediately struck by the diversity in form of both elicited and
spontaneous electrical activity in the single nerve cell. This applies
not only to the temporal patterns of all-or-none action potentials
but also to the graded somatic and dendritic potentials. The synaptic
membrane of a neuron, for example, is often found to be electrically
inexcitable and thus incapable of producing an action potential; yet
the graded, synaptically induced potentials show an amazing diversity
in form. In response to a presynaptic impulse, the postsynaptic
membrane may become hyperpolarized (inhibitory postsynaptic potential),
depolarized (excitatory postsynaptic potential), or remain at the
resting potential but with an increased permeability to certain ions
(a form of inhibition). The form of the postsynaptic potential in
response to an isolated presynaptic spike may vary from synapse to
synapse in several ways, as shown in Figure 1. Following a presynaptic
spike, the postsynaptic potential typically rises with some delay to
a peak value and then falls back toward the equilibrium or resting
potential. Three potentially important factors are the delay time
(synaptic delay), the peak amplitude (spatial weighting of synapse),
and the rate of fall toward the equilibrium potential (temporal
weighting of synapse). The responses of a synapse to individual spikes
in a volley may be progressively enhanced (facilitation), diminished
(antifacilitation), or neither (1), (2), (7), (8). Facilitation may be
in the form of progressively increased peak amplitude, or in the form
of progressively decreased rate of fall (see Figure 2). The time course
and magnitude of facilitation or antifacilitation may very well be
important synaptic parameters. In addition, the postsynaptic membrane
sometimes exhibits excitatory or inhibitory aftereffects (or both) on
cessation of a volley of presynaptic spikes (2), (7); and the time
course and magnitude of the aftereffects may be important parameters.
Clearly, even if one considers the synaptic potentials alone, he is
faced with an impressive variety of responses. Examples of the various
types of postsynaptic responses may be found in the literature, but for
purposes of the present discussion the idealized wave forms in Figure 2
will demonstrate the diversity of electrical behavior with which one is
faced.
[Illustration: A. EXCITATORY POSTSYNAPTIC POTENTIAL FROM APLYSIA (SEE
REF. 2)]
[Illustration: B. EXCITATORY POSTSYNAPTIC POTENTIAL FROM PANULIRUS (SEE
REF. 1)]
[Illustration: C. EXCITATORY POSTSYNAPTIC POTENTIAL FROM MAMMALIAN
MOTONEURONE (SEE REF. 24)]
[Illustration: D. PRESYNAPTIC SPIKE
Figure 1—Excitatory postsynaptic potentials in response to a single
presynaptic spike]
[Illustration: A. TEMPORAL FACILITATION]
[Illustration: B. AMPLITUDE FACILITATION]
[Illustration: C. ANTIFACILITATION]
[Illustration: D. INHIBITORY POSTSYNAPTIC POTENTIALS EXHIBITING
BIPHASIC REBOUND]
[Illustration: E. PRESYNAPTIC SPIKE BURST
Figure 2—Idealized postsynaptic potentials]
In addition to synaptically induced potentials, low-frequency,
spontaneous potential fluctuations have been observed in many neurons
(2), (7), (9), (10), (11). These fluctuations, generally referred to
as pacemaker potentials, are usually rhythmic and may be undulatory
or more nearly saw-toothed in form. The depolarizing phase may be
accompanied by a spike, a volley of spikes, or no spikes at all.
Pacemaker frequencies have been noted from ten or more cycles per
second down to one cycle every ten seconds or more. Some idealized
pacemaker wave forms are shown in Figure 3.
[Illustration: A. PERIODIC BURSTS]
[Illustration: B. PACEMAKER POTENTIALS WITHOUT SPIKES]
[Illustration: C. PACEMAKER POTENTIALS WITH SINGLE SPIKES ON
DEPOLARIZING PHASE
Figure 3—Idealized pacemaker potentials]
[Illustration: A. FORM OF VOLTAGE STIMULI AND RESULTING MEMBRANE
POTENTIAL CHANGES.]
[Illustration: B. RESPONSE CURVE OF TYPICAL GRADED RESPONSE REGION.
Figure 4—Graded response]
Bullock (7), (10), (12), (13) has demonstrated the existence of
a third type of subthreshold response, which he calls the graded
response. While the postsynaptic membrane is quite often electrically
inexcitable, other regions of the somatic and dendritic membranes
appear to be moderately excitable. It is in these regions that Bullock
observes the graded response. If one applies a series of pulsed voltage
stimuli to the graded-response region, the observed responses would be
similar to those shown in Figure 4A. Plotting the peak response voltage
as a function of the stimulus voltage would result in a curve similar
to that in Figure 4B (see Ref. 3, page 4). For small values of input
voltage, the response curve is linear; the membrane is passive. As the
stimulus voltage is increased, however, the response becomes more and
more disproportionate. The membrane is actively amplifying the stimulus
potential. At even higher values of stimulus potential, the system
becomes regenerative; and a full action potential results. The peak
amplitude of the response depends on the duration of the stimulus as
well as on the amplitude. It also depends on the rate of application of
the stimulus voltage. If the stimulus potential is a voltage ramp, for
example, the response will depend on the slope of the ramp. If the rate
of rise is sufficiently low, the membrane will respond in a passive
manner to voltages much greater than the spike threshold for suddenly
applied voltages. In other words, the graded-response regions appear to
accommodate to slowly varying potentials.
In terms of functional operation, we can think of the synapse as a
transducer. The input to this transducer is a spike or series of spikes
in the presynaptic axon. The output is an accumulative, long-lasting
potential which in some way (perhaps not uniquely) represents the
pattern of presynaptic spikes. The pacemaker appears to perform the
function of a clock, producing periodic spikes or spike bursts or
producing periodic changes in the over-all excitability of the neuron.
The graded-response regions appear to act as nonlinear amplifiers and,
occasionally, spike initiators. The net result of this electrical
activity is transformed into a series of spikes which originate at
spike initiation sites and are propagated along axons to other neurons.
The electrical activity in the neuron described above is summarized in
the following outline (taken in part from Bullock (7)):
1. Synaptic Potentials
a. Excitatory or inhibitory
b. Facilitated, antifacilitated, or neither
c. With excitatory aftereffect, inhibitory aftereffect,
neither, or both
2. Pacemaker Potentials
a. Relaxation type, undulatory type, or none at all
b. Producing single spike, spike burst, or no spikes
c. Rhythmic or sporadic
3. Graded Response (rate sensitive)
4. Spike Initiation
THE MODERN IONIC HYPOTHESIS
Hodgkin, Huxley, and Katz (3) and Hodgkin and Huxley (14), (15), (16),
in 1952, published a series of papers describing detailed measurements
of voltage, current, and time relationships in the giant axon of the
squid (_Loligo_). Hodgkin and Huxley (17) consolidated and formalized
these data into a set of simultaneous differential equations describing
the hypothetical time course of events during spike generation and
propagation. The hypothetical system which these equations describe is
the basis of the Modern Ionic Hypothesis.
The system proposed by Hodgkin and Huxley is basically one of dynamic
opposition of ionic fluxes across the axon membrane. The membrane
itself forms the boundary between two liquid phases—the intracellular
fluid and the extracellular fluid. The intracellular fluid is rich in
potassium ions and immobile organic anions, while the extracellular
fluid contains an abundance of sodium ions and chloride ions. The
membrane is slightly permeable to the potassium, sodium, and chloride
ions; so these ions tend to diffuse across the membrane. When the
axon is inactive (not propagating a spike), the membrane is much more
permeable to chloride and potassium ions than it is to sodium ions.
In this state, in fact, sodium ions are actively transported from the
inside of the membrane to the outside at a rate just sufficient to
balance the inward leakage. The relative sodium ion concentrations on
both sides of the membrane are thus fixed by the active transport rate,
and the net sodium flux across the membrane is effectively zero. The
potassium ions, on the other hand, tend to move out of the cell; while
chloride ions tend to move into it. The inside of the cell thus becomes
negative with respect to the outside. When the potential across the
membrane is sufficient to balance the inward diffusion of chloride with
an equal outward drift, and the outward diffusion of potassium with an
inward drift (and possibly an inward active exchange), equilibrium is
established. The equilibrium potential is normally in the range of 60
to 65 millivolts.
The resting neural membrane is thus polarized, with the inside
approximately 60 millivolts negative with respect to the outside.
Most of the Hodgkin-Huxley data is based on measurements of the
transmembrane current in response to an imposed stepwise reduction
(depolarization) of membrane potential. By varying the external
ion concentrations, Hodgkin and Huxley were able to resolve the
transmembrane current into two “active” components, the potassium
ion current and the sodium ion current. They found that while the
membrane permeabilities to chloride and most other inorganic ions
were relatively constant, the permeabilities to both potassium and
sodium were strongly dependent on membrane potential. In response to a
suddenly applied (step) depolarization, the sodium permeability rises
rapidly to a peak and then declines exponentially to a steady value.
The potassium permeability, on the other hand, rises with considerable
delay to a value which is maintained as long as the membrane remains
depolarized. The magnitudes of both the potassium and the sodium
permeabilities increase monotonically with increasing depolarization.
A small imposed depolarization will result in an immediately
increased sodium permeability. The resulting increased influx of
sodium ions results in further depolarization; and the process
becomes regenerative, producing the all-or-none action potential.
At the peak of the action potential, the sodium conductance begins
to decline, while the delayed potassium conductance is increasing.
Recovery is brought about by an efflux of potassium ions, and both
ionic permeabilities fall rapidly as the membrane is repolarized.
The potassium permeability, however, falls less rapidly than that of
sodium. This is basically the explanation of the all-or-none spike
according to the Modern Ionic Hypothesis.
[Illustration: Figure 5—Hodgkin-Huxley representation of small area of
axon membrane]
[Illustration: Figure 6—Typical responses of sodium conductance and
potassium conductance to imposed step depolarization]
By defining the net driving force on any given ion species as the
difference between the membrane potential and the equilibrium potential
for that ion and describing permeability changes in terms of equivalent
electrical conductance changes, Hodgkin and Huxley reduced the ionic
model to the electrical equivalent in Figure 5. The important dynamic
variables in this equivalent network are the sodium conductance
(G{Na}) and the potassium conductance (G{K}). The change in the sodium
conductance in response to a step depolarization is shown in Figure 6B.
This change can be characterized by seven voltage dependent parameters:
1. Delay time—generally much less than 1 msec
2. Rise time—1 msec or less
3. Magnitude of peak conductance—increases
monotonically with increasing depolarization
4. Inactivation time constant—decreases
monotonically with increasing depolarization.
5. Time constant of recovery from
inactivation—incomplete data
6. Magnitude of steady-state conductance—increases
monotonically with increasing depolarization
7. Fall time on sudden repolarization—less than 1 msec.
Figure 6B shows the potassium conductance change in response to
an imposed step depolarization. Four parameters are sufficient to
characterize this response:
1. Delay time—decreases monotonically with
increasing depolarization
2. Rise time—decreases monotonically with increasing
depolarization
3. Magnitude of steady-state conductance—increases
monotonically with increasing depolarization
4. Fall time on sudden repolarization—8 msec
or more, decreases slightly with increasing
depolarization.
In addition to the aforementioned parameters, the transient portion of
the sodium conductance appears to exhibit an accommodation to slowly
varying membrane potentials. The time constants of accommodation appear
to be those of inactivation or recovery from inactivation—depending on
the direction of change in the membrane potential (18). The remaining
elements in the Hodgkin-Huxley model are constant and are listed below:
1. Potassium potential—80 to 85 mv (inside negative)
2. Sodium potential—45 to 50 mv (inside positive)
3. Leakage potential—38 to 43 mv (inside negative)
4. Leakage conductance—approx. 0.23 millimhos/cm²
5. Membrane capacitance—approx. 1 μf/cm²
6. Resting potential—60 to 65 mv
7. Spike amplitude—approx. 100 mv
ELECTRONIC SIMULATION OF THE HODGKIN-HUXLEY MODEL
[Illustration: Figure 7—System diagram for electronic simulation of the
Hodgkin-Huxley model]
Given a suitable means of generating the conductance functions,
G_{Na}(v,t) and G_{K}(v,t), one can readily stimulate the essential
aspects of the Modern Ionic Hypothesis. If we wish to do this
electronically, we have two problems. First, we must synthesize
a network whose input is the membrane potential and whose output
is a voltage or current proportional to the desired conductance
function. Second, we must transform the output from a voltage or
current to an effective electronic conductance. The former implies
the need for nonlinear, active filters, while the latter implies
the need for multipliers. The basic block diagram is shown in
Figure 7. Several distinct realizations of this system have been
developed in our laboratory, and in each case the results were the
same. With parameters adjusted to closely match the data of Hodgkin
and Huxley, the electronic model exhibits all of the important
properties of the axon. It produces spikes of 1 to 2 msec duration
with a threshold of approximately 5% to 10% of the spike amplitude.
The applied stimulus is generally followed by a prepotential, then
an active rise of less than 1 msec, followed by an active recovery.
The after-depolarization generally lasts several msec, followed by
a prolonged after-hyperpolarization. The model exhibits the typical
strength-duration curve, with rheobase of 5% to 10% of the spike
amplitude. For sufficiently prolonged sodium inactivation (long time
constant of recovery from inactivation), the model also exhibits an
effect identical to classical Wedensky inhibition (18). Thus, as would
be expected, the electronic model simulates very well the electrical
properties of the axon.
In addition to the axon properties, however, the electronic model is
able to reproduce all of the somatic and dendritic activity outlined
in the section on subthreshold activity. Simulation of the pacemaker
and graded-response potentials is accomplished without additional
circuitry. In the case of synaptically induced potentials, however,
auxiliary networks are required. These networks provide additive terms
to the variable conductances in accordance with current notions on
synaptic transmission (19). Two types of networks have been used. In
both, the inputs are simulated presynaptic spikes, and in both the
outputs are the resulting simulated chemical transmitter concentration.
In both, the transmitter substance was assumed to be injected at a
constant rate during a presynaptic spike and subsequently inactivated
in the presence of an enzyme. One network simulates a first-order
chemical reaction, where the enzyme concentration is effectively
constant. The other simulates a second-order chemical reaction,
where the enzyme concentration is assumed to be reduced during the
inactivation process. For simulation of an excitatory synapse, the
output of the auxiliary network is added directly to G_{Na} in the
electronic model. For inhibition, it is added to G_{K}. With the
parameters of the electronic membrane model set at the values measured
by Hodgkin and Huxley, we have attempted to simulate synaptic activity
with the aid of the two types of auxiliary networks. In the case of
the simulated first-order reaction, the excitatory synapse exhibits
facilitation, antifacilitation, or neither—depending on the setting
of a single parameter, the transmitter inactivation rate (_i.e._,
the effective enzyme concentration). This parameter would appear,
in passing, to be one of the most probable synaptic variables. In
this case, the mechanisms for facilitation and antifacilitation are
contained in the simulated postsynaptic membrane. Facilitation is due
to the nonlinear dependence of G_{Na} on membrane potential, while
antifacilitation is due to inactivation of G_{Na}. The occurrence
of one form of response or the other is determined by the relative
importance of the two mechanisms (18). Grundfest (20) has mentioned
both of these mechanisms as potentially facilitory and antifacilitory,
respectively. The simulated inhibitory synapse with the first order
input is capable of facilitation (18), but no antifacilitation has been
observed. Again, the presence or absence of facilitation is determined
by the inactivation rate.
With the simulated second-order reaction, both excitatory and
inhibitory synapses exhibit facilitation. In this case, two facilitory
mechanisms are present—one in the postsynaptic membrane and one in the
nonconstant transmitter inactivation reaction. The active membrane
currents can, in fact, be removed; and this system will still exhibit
facilitation. With the second-order auxiliary network, the presence
of excitatory facilitation, antifacilitation, or neither depends
on the initial, or resting, transmitter inactivation rate. The
synaptic behavior also depends parametrically on the simulated enzyme
reactivation rate. Inhibitory antifacilitation can be introduced with
either type of auxiliary network by limiting the simulated presynaptic
transmitter supply.
Certain classes of aftereffects are inherent in the mechanisms of the
Ionic Hypothesis. In the electronic model, aftereffects are observed
following presynaptic volleys with either type of auxiliary network.
Following a volley of spikes into the simulated excitatory synapse,
for example, rebound hyperpolarization may or may not occur depending
on the simulated transmitter inactivation rate. If the inactivation
rate is sufficiently high, rebound will occur. This rebound can be
monophasic (inhibitory phase only) or polyphasic (successive cycles
of excitation and inhibition). Following a volley of spikes into the
simulated inhibitory synapse, rebound depolarization may or may not
occur depending on the simulated transmitter inactivation rate. This
rebound can also be monophasic or polyphasic. Sustained postexcitatory
depolarization and sustained postinhibitory hyperpolarization (2) have
been achieved in the model by making the transmitter inactivation rate
sufficiently low.
The general forms of the postsynaptic potentials simulated with
the electronic model are strikingly similar to those published in
the literature for real neurons. The first-order auxiliary network
produces facilitation of a form almost identical to that shown by Otani
and Bullock (8) while the second-order auxiliary network produces
facilitation of the type shown by Chalazonitis and Arvanitake (2).
The excitatory antifacilitation is almost identical to that shown by
Hagiwara and Bullock (1) in both form and dependence on presynaptic
spike frequency. In every case, the synaptic behavior is determined
by the effective rate of transmitter inactivation, which in real
neurons would presumably be directly proportional to the effective
concentration of inactivating enzyme at the synapse.
Pacemaker potentials are easily simulated with the electronic model
without the use of auxiliary networks. This is achieved either by
inserting a large, variable shunt resistor across the simulated
membrane (see Figure 5) or by allowing a small sodium current leakage
at the resting potential. With the remaining parameters of the
model set as close as possible to the values determined by Hodgkin
and Huxley, the leakage current induces low-frequency, spontaneous
spiking. The spike frequency increases monotonically with increasing
leakage current. In addition, if the sodium conductance inactivation
is allowed to accumulate over several spikes, periodic spike pairs
and spike bursts will result. Subthreshold pacemaker potentials have
also been observed in the model, but with parameter values set close
to the Hodgkin-Huxley data these are generally higher in frequency
than pacemaker potentials in real neurons. It is interesting that
a pacemaker mode may exist in the absence of the simulated sodium
conductance. It is a very high-frequency mode (50 cps or more)
and results from the alternating dominance of potassium current
and chloride (or leakage ion) current in determining the membrane
potential. The significance of this mode cannot be assessed until
better data is available for the potassium conductance at low levels
of depolarization in real neurons. In general, as far as the model is
concerned, pacemaker potentials are possible because the potassium
conductance is delayed in both its rise with depolarization and its
fall with repolarization.
Rate sensitive graded response has also been observed in the electronic
model. The rate sensitivity—or accommodation—is due to the sodium
conductance inactivation. The response of the model to an imposed ramp
depolarization was discussed in Reference 18. At this time, several
alternative model parameters could be altered to bring about reduced
electrical excitability. None of the parameter changes was very
satisfying, however, because none of them was in any way justified by
physiological data. We have since found that the membrane capacitance,
a plausible parameter in view of recent physiological findings, can
completely determine the electrical excitability. Thus, with the
capacitance determined by Hodgkin and Huxley (1 microfarad per cm²),
the model exhibits excitability characteristic of the axon. As the
capacitance is increased, the model becomes less excitable until, with
10 or 12 μμf, it is effectively inexcitable. Thus, with an increased
capacitance—but with all the remaining parameters set as close as
possible to the Hodgkin-Huxley values—the electronic model exhibits the
characteristics of Bullock’s graded-response regions.
Whether membrane capacitance is the determining factor in real neurons
is, of course, a matter of speculation. Quite a controversy is raging
over membrane capacity measurements (see Rall (21)), but the evidence
indicates that the capacity in the soma is considerably greater than
that in the axon (6), (22).
It should be added that increasing the capacitance until the membrane
model becomes inexcitable has little effect on the variety of available
simulated synaptic responses. Facilitation, antifacilitation, and
rebound are still present and still depend on the transmitter
inactivation rate. Thus, in the model, we can have a truly inexcitable
membrane which nevertheless utilizes the active membrane conductances
to provide facilitation or antifacilitation, and rebound. The simulated
subthreshold pacemaker potentials are much more realistic with the
increased capacitance, being lower in frequency and more natural in
form.
In one case, the electronic model predicted behavior which was
subsequently reported in real neurons. This was in respect to the
interaction of synaptic potentials and pacemaker potential. It was
noted in early experiments that when the model was set in a pacemaker
mode, and periodic spikes were applied to the simulated inhibitory
synapse, the pacemaker frequency could be modified; and, in fact,
it would tend to lock on to the stimulus frequency. This produced
a paradoxical effect whereby the frequency of spontaneous spikes
was actually increased by increasing the frequency of inhibitory
synaptic stimuli. At very low stimulus frequencies, the spontaneous
pacemaker frequency was not appreciably perturbed. As the stimulus
frequency was increased, and approached the basic pacemaker frequency,
the latter tended to lock on and follow further increases in the
stimulus frequency. When the stimulus frequency became too high for
the pacemaker to follow, the latter decreased abruptly in frequency
and locked on to the first subharmonic. As the stimulus frequency was
further increased, the pacemaker frequency would increase, then skip to
the next harmonic, then increase again, _etc._ This type of behavior
was observed by Moore _et al._ (23) in _Aplysia_ and reported at the
San Diego Symposium for Biomedical Electronics shortly after it was
observed by the author in the electronic model.
Thus, we have shown that an electronic analog with all parameters
except membrane capacitance fixed at values close to those of Hodgkin
and Huxley, can provide all of the normal threshold or axonal
behavior and also all of the subthreshold somatic and dendritic
behavior outlined on page 7. Whether or not this is of physiological
significance, it certainly provides a unifying basis for construction
of electronic neural analogs. Simple circuits, based on the
Hodgkin-Huxley model and providing all of the aforementioned behavior,
have been constructed with ten or fewer inexpensive transistors with
a normal complement of associated circuitry (18). In the near future
we hope to utilize several models of this type to help assess the
information-processing capabilities not only of individual neurons but
also of small groups or networks of neurons.
REFERENCES
1. Hagiwara, S., and Bullock, T. H.
“Intracellular Potentials in Pacemaker and Integrative Neurons of
the Lobster Cardiac Ganglion,”
_J. Cell and Comp. Physiol._ =50 (No. 1)=:25-48 (1957)
2. Chalazonitis, N., and Arvanitaki, A.,
“Slow Changes during and following Repetitive Synaptic Activation
in Ganglion Nerve Cells,”
_Bull. Inst. Oceanogr. Monaco_ =No. 1225=:1-23 (1961)
3. Hodgkin, A. L., Huxley, A. F., and Katz, B.,
“Measurement of Current-Voltage Relations in the Membrane of the
Giant Axon of _Loligo_,”
_J. Physiol._ =116=:424-448 (1952)
4. Hagiwara, S., and Saito, N.,
“Voltage-Current Relations in Nerve Cell Membrane of Onchidium
_verruculatum_,”
_J. Physiol._ =148=:161-179 (1959)
5. Hagiwara, S., and Saito, N.,
“Membrane Potential Change and Membrane Current in Supramedullary
Nerve Cell of Puffer,”
_J. Neurophysiol._ =22=:204-221 (1959)
6. Hagiwara, S.,
“Current-Voltage Relations of Nerve Cell Membrane,”
“Electrical Activity of Single Cells,”
Igakushoin, Hongo, Tokyo (1960)
7. Bullock, T. H.,
“Parameters of Integrative Action of the Nervous System at the
Neuronal Level,”
_Experimental Cell Research Suppl._ =5=:323-337 (1958)
8. Otani, T., and Bullock, T. H.,
“Effects of Presetting the Membrane Potential of the Soma of
Spontaneous and Integrating Ganglion Cells,”
_Physiological Zoology_ =32 (No. 2)=:104-114 (1959)
9. Bullock, T. H., and Terzuolo, C. A.,
“Diverse Forms of Activity in the Somata of Spontaneous and
Integrating Ganglion Cells,”
_J. Physiol._ =138=:343-364 (1957)
10. Bullock, T. H.,
“Neuron Doctrine and Electrophysiology,”
_Science_ =129 (No. 3355)=:997-1002 (1959)
11. Chalazonitis, N., and Arvanitaki, A.,
“Slow Waves and Associated Spiking in Nerve Cells of
_Aplysia_,”
_Bull. Inst. Oceanogr. Monaco_ =No. 1224=:1-15 (1961)
12. Bullock, T. H.,
“Properties of a Single Synapse in the Stellate Ganglion of
Squid,”
_J. Neurophysiol._ =11=:343-364 (1948)
13. Bullock, T. H.,
“Neuronal Integrative Mechanisms,”
“Recent Advances in Invertebrate Physiology,”
Scheer, B. T., ed., Eugene, Oregon:Univ. Oregon Press 1957
14. Hodgkin, A. L., and Huxley, A. F.,
“Currents Carried by Sodium and Potassium Ions through the
Membrane of the Giant Axon of Loligo,”
_J. Physiol._ =116=:449-472 (1952)
15. Hodgkin, A. L., and Huxley, A. F.,
“The Components of Membrane Conductance in the Giant Axon of
_Loligo_,”
_J. Physiol._ =116=:473-496 (1952)
16. Hodgkin, A. L., and Huxley, A. F.,
“The Dual Effect of Membrane Potential on Sodium Conductance in
the Giant Axon of _Loligo_,”
_J. Physiol._ =116=:497-506 (1952)
17. Hodgkin, A. L., and Huxley, A. F.,
“A Quantitative Description of Membrane Current and its
Application to Conduction and Excitation in Nerve,”
_J. Physiol._ =117=:500-544 (1952)
18. Lewis, E. R.,
“An Electronic Analog of the Neuron Based on the Dynamics of
Potassium and Sodium Ion Fluxes,”
“Neural Theory and Modeling,”
R. F. Reiss, ed., Palo Alto, California:Stanford University
Press, 1964
19. Eccles, J. C.,
_Physiology of Synapses_,
Berlin:Springer-Verlag, 1963
20. Grundfest, H.,
“Excitation Triggers in Post-Junctional Cells,”
“Physiological Triggers,”
T. H. Bullock, ed., Washington, D.C.:American Physiological
Society, 1955
21. Rall, W.,
“Membrane Potential Transients and Membrane Time Constants of
Motoneurons,”
_Exp. Neurol._ =2=:503-532 (1960)
22. Araki, T., and Otani, T.,
“The Response of Single Motoneurones to Direct Stimulation,”
_J. Neurophysiol._ =18=:472-485 (1955)
23. Moore, G. P., Perkel, D. H., and Segundo, J. P.,
“Stability Patterns in Interneuronal Pacemaker Regulation,”
_Proceedings of the San Diego Symposium for Biomedical
Engineering_, San Diego, California, 1963
24. Eccles, J. C.,
_The Neurophysiological Basis of Mind_,
Oxford:Clarendon Press, 1952
Fields and Waves in Excitable Cellular Structures
R. M. STEWART
_Space General Corporation
El Monte, California_
“Study of living processes by the physiological
method only proceeded laboriously behind the study of
non-living systems. Knowledge about respiration, for
instance, began to become well organized as the study
of combustion proceeded, since this is an analogous
operation....”
J. Z. Young (24)
INTRODUCTION
The study of electrical fields in densely-packed cellular media is
prompted primarily by a desire to understand more fully the details
of brain mechanism and its relation to behavior. Our work has
specifically been directed toward an attempt to model such structures
and mechanisms, using relatively simple inorganic materials.
The prototype for such experiments is the “Lillie[1] iron-wire nerve
model.” Over a hundred years ago, it had been observed that visible
waves were produced on the surface of a piece of iron submerged in
nitric acid when and where the iron is touched by a piece of zinc.
After a short period of apparent fatigue, the wire recovers and can
again support a wave when stimulated. Major support for the idea that
such impulses are in fact directly related to peripheral nerve impulses
came from Lillie around 1920. Along an entirely different line,
various persons have noted the morphological and dynamic similarity of
dendrites in brain and those which sometimes grow by electrodeposition
of metals from solution. Gordon Pask (17), especially, has pointed to
this similarity and has discussed in a general way the concomitant
possibility of a physical model for the persistent memory trace.
[1] For review articles see: Lillie (13), Franck (6).
By combining and extending such concepts and techniques, we hope to
produce a macroscopic model of “gray matter,” the structural matrix of
which will consist of a dense, homogeneously-mixed, conglomerate of
small pellets, capable of supporting internal waves of excitation, of
changing electrical behavior through internal fine-structure growth,
and of forming temporal associations in response to peripheral shocks.
A few experimenters have subsequently pursued the iron-wire
nerve-impulse analogy further, hoping thereby to illuminate the
mechanisms of nerve excitation, impulse transmission and recovery,
but interest has generally been quite low. It has remained fairly
undisturbed in the text books and lecture demonstrations of medical
students, as a picturesque aid to their formal education. On the
outer fringes of biology, still less interest has been displayed;
the philosophical vitalists would surely be revolted by the idea of
such models of mind and memory, and at the other end of the scale,
contemporary computer engineers generally assume that a nerve cell
operates much too slowly to be of any value. This lack of interest
is certainly due, in part, to success in developing techniques of
monitoring individual nerve fibers directly to the point that it is
just about as easy to work with large nerve fibers (and even peripheral
and spinal junctions) as it is to work with iron wires. Under such
circumstances, the model has only limited value, perhaps just to the
extent that it emphasizes the role of factors other than specific
molecular structure and local chemical reactions in the dynamics of
nerve action.
When we leave the questions of impulse transmission on long fibers
and peripheral junctions, however, and attempt to discuss the brain,
there can be hardly any doubt that the development of a meaningful
physical model technique would be of great value. Brain tissue is
soft and sensitive, the cellular structures are small, tangled, and
incredibly numerous. Therefore (Young (24)), “ ... physiologists hope
that after having learned a lot about nerve-impulses in the nerves they
will be able to go on to study how these impulses interact when they
reach the brain. [But], we must not assume that we shall understand
the brain only in the terms we have learned to use for the nerves.
The function of nerves is to carry impulses—like telegraph wires. The
functions of brains is something else.” But, confronted with such
awesome experimental difficulties, with no comprehensive mathematical
theory in sight, we are largely limited otherwise to verbal discourses,
rationales and theorizing, a hopelessly clumsy tool for the development
of an adequate understanding of brain function. A little over ten years
ago Sperry (19) said, “Present day science is quite at a loss even
to begin to describe the neural events involved in the simplest form
of mental activity.” This situation has not changed much today. The
development, study, and understanding of complex high-density cellular
structures which incorporate characteristics of both the Lillie and
Pask models may, it is hoped, alleviate this situation. There would
also be fairly obvious technological applications for such techniques
if highly developed and which, more than any other consideration, has
prompted support for this work.
Experiments to date have been devised which demonstrate the following
basic physical functional characteristics:
(1) Control of bulk resistivity of electrolytes containing
closely-packed, poorly-conducting pellets
(2) Circulation of regenerative waves on closed loops
(3) Strong coupling between isolated excitable sites
(4) Logically-complete wave interactions, including facilitation
and annihilation
(5) Dendrite growth by electrodeposition in “closed” excitable
systems
(6) Subthreshold distributed field effects, especially in
locally-refractory regions.
In addition, our attention has necessarily been directed to various
problems of general experimental technique and choice of materials,
especially as related to stability, fast recovery and long life.
However, in order to understand the possible significance of, and
motivation for such experiments, some related modern concepts of
neurophysiology, histology and psychology will be reviewed very
briefly. These concepts are, respectively:
(1) Cellular structure in the central nervous system
(2) Short-term or “ephemeral” memory
(3) The synapse
(4) Inhibition
(5) Long-term memory traces or engram
(6) Spatially-diffuse temporal association and learning.
SOME CONTEMPORARY CONCEPTS
Since we are attempting to duplicate processes other than chemical,
per se, we will forego any reference to the extensive literature of
neurochemistry. It should not be surprising though if, at the neglect
of the fundamental biological processes of growth, reproduction and
metabolism, it proves possible to imitate some learning mechanisms
with grossly less complex molecular structures. There is also
much talk of chemical versus electrical theories and mechanisms in
neurophysiology. The distinction, when it can be made, seems to hinge
on the question of the scale of size of significant interactions. Thus,
“chemical” interactions presumably take place at molecular distances,
possibly as a result of or subsequent to a certain amount of thermal
diffusion. “Electrical” interactions, on the other hand, are generally
understood to imply longer range or larger scale macroscopic fields.
1. Cellular Structure
The human brain contains approximately 10¹⁰ neurons to which the
neuron theory assigns the primary role in central nervous activity.
These cells occupy, however, a relatively small fraction of the total
volume. There are, for example, approximately 10 times that number of
neuroglia, cells of relatively indeterminate function. Each neuron
(consisting of cell body, dendrites and, sometimes, an axon) comes into
close contact with the dendrites of other neurones at some thousands
of places, these synapses and “ephapses” being spaced approximately 5μ
apart (1). The total number of such apparent junctions is therefore
of the order of 10¹³. In spite of infinite fine-structure variations
when viewed with slightly blurred vision, the cellular structure of
the brain is remarkably homogeneous. In the cortex, at least, the
extensions of most cells are relatively short, and when the cortex is
at rest, it appears from the large EEG alpha-rhythms that large numbers
of cells beat together in unison. Quoting again from Sperry, “In short,
current brain theory encourages us to try to correlate our subjective
psychic experience with the activity of relatively homogeneous nerve
cell units conducting essentially homogeneous impulses, through roughly
homogeneous cerebral tissue.”
2. Short-Term Memory
A train of impulses simply travelling on a long fiber may, for
example, be regarded as a short-term memory much in the same way as
a delay line acts as a transient memory in a computer. A similar
but slightly longer term memory may also be thought of to exist in
the form of waves circulating in closed loops (23). In fact, it is
almost universally held today that most significant memory occurs
in two basic interrelated ways. First of all, such a short-term
circulating, reverberatory or regenerative memory which, however, could
not conceivably persist through such things as coma, anesthesia,
concussion, extreme cold, deep sleep and convulsive seizures and
thus, secondly, a long-term memory trace which must somehow reside
in a semipermanent fine-structural change. As Hebb (9) stated, “A
reverbratory trace might cooperate with a structural change and carry
the memory until the growth change is made.”
3. The Synapse
The current most highly regarded specific conception of the synapse
is largely due to and has been best described by Eccles (5): “ ...
the synaptic connections between nerve cells are the only functional
connections of any significance. These synapses are of two types,
excitatory and inhibitory, the former type tending to make nerve cells
discharge impulses, the other to suppress the discharge. There is now
convincing evidence that in vertebrate synapses each type operates
through specific chemical transmitter substances ...”. In response to
a presentation by Hebb (10), Eccles was quoted as saying, “One final
point, and that is if there is electrical interaction, and we have seen
from Dr. Estable’s work the complexity of connections, and we now know
from the electronmicroscopists that there is no free space, only 200
Å clefts, everywhere in the central nervous system, then everything
should be electrically interacted with everything else. I think this is
only electrical background noise and, that when we lift with specific
chemical connections above that noise we get a significant operational
system. I would say that there is electrical interaction but it is just
a noise, a nuisance.” Eccles’ conclusions are primarily based on data
obtained in the peripheral nervous system and the spinal cord. But
there is overwhelming reason to expect that cellular interactions in
the brain are an entirely different affair. For example, “The highest
centres in the octopus, as in vertebrates and arthropods, contain many
small neurons. This finding is such a commonplace, that we have perhaps
failed in the past to make the fullest inquiry into its implications.
Many of these small cells possess numerous processes, but no axon. It
is difficult to see, therefore, that their function can be conductive
in the ordinary sense. Most of our ideas about nervous functioning are
based on the assumption that each neuron acts essentially as a link in
some chain of conduction, but there is really no warrant for this in
the case of cells with many short branches. Until we know more of the
relations of these processes to each other in the neuropile it would
be unwise to say more. It is possible that the effective part of the
discharge of such cells is not as it is in conduction in long pathways,
the internal circuit that returns through the same fiber, but the
external circuit that enters other processes, ...” (3).
4. Inhibition
The inhibitory chemical transmitter substance postulated by Eccles
has never been detected in spite of numerous efforts to do so. The
mechanism(s) of inhibition is perhaps the key to the question of
cellular interaction and, in one form or another, must be accounted for
in any adequate theory.
Other rather specific forms of excitation and inhibition interaction
have been proposed at one time or another. Perhaps the best example is
the polar neuron of Gesell (8) and, more recently, Retzlaff (18). In
such a concept, excitatory and inhibitory couplings differ basically
because of a macroscopic structural difference at the cellular level;
that is, various arrangements or orientation of intimate cellular
structures give rise to either excitation or inhibition.
5. Long-Term Memory
Most modern theories of semipermanent structural change (or _engrams_,
as they are sometimes called) look either to the molecular level or to
the cellular level. Various specific locales for the engram have been
suggested, including (1) modifications of RNA molecular structure,
(2) changes of cell size, synapse area or dendrite extensions, (3)
neuropile modification, and (4) local changes in the cell membrane.
There is, in fact, rather direct evidence of the growth of neurons or
their dendrites with use and the diminution or atrophy of dendrites
with disuse. The apical dendrite of pyramidal neurones becomes thicker
and more twisted with continuing activity, nerve fibers swell when
active, sprout additional branches (at least in the spinal cord) and
presumably increase the size and number of their terminal knobs.
As pointed out by Konorski (11), the morphological conception of
plasticity according to which plastic changes would be related to the
formation and multiplication of new synaptic junctions goes back at
least as far as Ramon y Cajal in 1904. Whatever the substrate of the
memory trace, it is, at least in adults, remarkably immune to extensive
brain damage and as Young (24) has said: “ ... this question of the
nature of the memory trace is one of the most obscure and disputed in
the whole of biology.”
6. Field Effects and Learning
First, from Boycott and Young (3), “The current conception, on which
most discussions of learning still concentrate, is that the nervous
system consists essentially of an aggregate of chains of conductors,
linked at key points by synapses. This reflex conception, springing
probably from Cartesian theory and method, has no doubt proved of
outstanding value in helping us to analyse the actions of the spinal
cord, but it can be argued that it has actually obstructed the
development of understanding of cerebral function.”
Most observable evidence of learning and memory is extremely complex
and its interpretation full of traps. Learning in its broadest sense
might be detected as a semipermanent change of behavior pattern brought
about as a result of experience. Within that kind of definition, we
can surely identify several distinctly different types of learning,
presumably with distinctly different kinds of mechanisms associated
with each one. But, if we are to stick by our definition of a condition
of semipermanent change of behavior as a criterion for learning, then
we may also be misled into considering the development of a neurosis,
for example, as learning, or even a deep coma as learning.
When we come to consider field effects, current theories tend to get
fairly obscure, but there seems to be an almost universal recognition
of the fact that such fields are significant. For example, Morrell
(16) says in his review of electrophysiological contributions to the
neural basis of learning, “A growing body of knowledge (see reviews
by Purpura, Grundfest, and Bishop) suggests that the most significant
integrative work of the central nervous system is carried on in graded
response elements—elements in which the degree of reaction depends upon
stimulus intensity and is not all-or-none, which have no refractory
period and in which continuously varying potential changes of either
sign occur and mix and algebraically sum.” Gerard (7) also makes a
number of general comments along these lines. “These attributes of
a given cell are, in turn, normally controlled by impulses arising
from other regions, by fields surrounding them—both electric and
chemical—electric and chemical fields can strongly influence the
interaction of neurones. This has been amply expounded in the case of
the electric fields.”
Learning situations involving “punishment” and “reward” or,
subjectively, “pain” and “pleasure” may very likely be associated
with transient but structurally widespread field effects. States of
distress and of success seem to exert a lasting influence on behavior
only in relation to _simultaneous_ sensory events or, better yet,
sensory events just immediately _preceding_ in time. For example, the
“anticipatory” nature of a conditioned reflex has been widely noted
(21). From a structural point of view, it is as if recently active
sites regardless of location or function were especially sensitive to
extensive fields. There is a known inherent electrical property of both
nerve membrane and passive iron surface that could hold the answer to
this mechanism of spatially-diffuse temporal association; namely, the
surface resistance drops to less than 1 per cent of its resting value
during the refractory period which immediately follows activation.
EXPERIMENTAL TECHNIQUE
In almost all experiments, the basic signal-energy mechanism employed
has been essentially that one studied most extensively by Lillie (12),
Bonhoeffer (2), Yamagiwa (22), Matumoto and Goto (14) and others,
_i.e._, activation, impulse propagation and recovery on the normally
passive surface of a piece of iron immersed in nitric acid or of
cobalt in chromic acid (20). The iron we have used most frequently
is of about 99.99% purity, which gives performance more consistent
than but similar to that obtained using cleaned “coat-hanger” wires.
The acid used most frequently by us is about 53-55% aqueous solution
by weight, substantially more dilute than that predominantly used by
previous investigators. The most frequently reported concentration has
been 68-70%, a solution which is quite stable and, hence, much easier
to work with in open containers than the weaker solutions, results in
very fast waves but gives, at room temperatures, a very long refractory
period (typically, 15 minutes). A noble metal (such as silver, gold
or platinum) placed in contact with the surface of the iron has a
stabilizing effect (14) presumably through the action of local currents
and provides a simple and useful technique whereby, with dilution,
both stability and fast recovery (1 second) can be achieved in simple
demonstrations and experiments.
Experiments involving the growth by electrodeposition and study of
metallic dendrites are done with an eye toward electrical, physical
and chemical compatibility with the energy-producing system outlined
above. Best results to date (from the standpoints of stability,
non-reactivity, and morphological similarity to neurological
structures) have been obtained by dissolving various amounts of gold
chloride salt in 53-55% HNO₃.
An apparatus has been devised and assembled for the purpose of
containing and controlling our primary experiments. (See Figure 1).
Its two major components are a test chamber (on the left in Figure
1) and a fluid exchanger (on the right). In normal operation the
test chamber, which is very rigid and well sealed after placing the
experimental assembly inside, is completely filled with electrolyte
(or, initially, an inert fluid) to the exclusion of all air pockets and
bubbles. Thus encapsulated, it is possible to perform experiments which
would otherwise be impossible due to instability. The instability which
plagues such experiments is manifested in copious generation of bubbles
on and subsequent rapid disintegration of all “excitable” material
(_i.e._, iron). Preliminary experiments indicated that such “bubble
instability” could be suppressed by constraining the volume available
to expansion. In particular, response and recovery times can now be
decreased substantially and work can proceed with complex systems of
interest such as aggregates containing many small iron pellets.
The test chamber is provided with a heater (and thermostatic control)
which makes possible electrochemical impulse response and recovery
times comparable to those of the nervous system (1 to 10 msec). The
fluid-exchanger is so arranged that fluid in the test chamber can be
arbitrarily changed or renewed by exchange within a rigid, sealed,
completely liquid-filled (“isochoric”) loop. Thus, stability can
be maintained for long periods of time and over a wide variety of
investigative or operating conditions.
Most of the parts of this apparatus are made of stainless steel and
are sealed with polyethylene and teflon. There is a small quartz
observation window on the test chamber, two small lighting ports, a
pressure transducer, thermocouple, screw-and-piston pressure actuator
and umbilical connector for experimental electrical inputs and outputs.
BASIC EXPERIMENTS
The basic types of experiments described in the following sections
are numbered for comparison to correspond roughly to related
neurophysiological concepts summarized in the previous section.
1. Cellular Structure
The primary object of our research is the control and determination of
dynamic behavior in response to electrical stimulation in close-packed
aggregates of small pellets submerged in electrolyte. Typically, the
aggregate contains (among other things) iron and the electrolyte
contains nitric acid, this combination making possible the propagation
of electrochemical surface waves of excitation through the body of
the aggregate similar to those of the Lillie iron-wire nerve model.
The iron pellets are imbedded in and supported by a matrix of small
dielectric (such as glass) pellets. Furthermore, with the addition
of soluble salts of various noble metals to the electrolyte, long
interstitial dendritic or fibrous structures of the second metal can
be formed whose length and distribution change by electrodeposition in
response to either internal or externally generated fields.
[Illustration: Figure 1—Test chamber and fluid exchanger]
Coupling between isolated excitable (iron) sites is greatly affected
by the fine structure and effective bulk resistivity of the glass and
fluid medium which supports and fills the space between such sites.
In general (see Section 3, following) it is necessary, to promote
strong coupling between small structures, to impede the “short-circuit”
return flow of current from an active or excited surface, through
the electrolyte and back through the dendritic structure attached
to the same excitable site. This calls for control (increase) of
the bulk resistivity, preferably by means specifically independent
of electrolyte composition, which relates to and affects surface
phenomena such as recovery (_i.e._, the “refractory” period). Figure 2
illustrates the way in which this is being done, _i.e._, by appropriate
choice of particle size distributions. The case illustrated shows
the approximate proper volume ratios for maximum resistivity in a
two-size-phase random mixture of spheres.
2. Regenerative Loops
Figure 3 shows an iron loop (about 2-inch diameter) wrapped with a
silver wire helix which is quite stable in 53-55% acid and which
will easily support a circulating pattern of three impulses. For
demonstration, unilateral waves can be generated by first touching the
iron with a piece of zinc (which produces two oppositely travelling
waves) and then blocking one of them with a piece of platinum or a
small platinum screen attached to the end of a stick or wand. Carbon
blocks may also be used for this purpose.
The smallest regenerative or reverberatory loop which we are at present
able to devise is about 1 mm in diameter. Multiple waves, as expected,
produce stable patterns in which all impulses are equally spaced. This
phenomenon can be related to the slightly slower speed characteristic
of the relative refractory period as compared with a more fully
recovered zone.
[Illustration]
[Illustration: Figure 2—Conductivity control—mixed pellet-size
aggregates]
[Illustration: Figure 3—Regenerative or reverberatory loop]
3. Strong Coupling
If two touching pieces of iron are placed in a bath of nitric acid, a
wave generated on one will ordinarily spread to the other. As is to be
expected, a similar result is obtained if the two pieces are connected
through an external conducting wire. However, if they are isolated,
strong coupling does not ordinarily occur, especially if the elements
are small in comparison with a “critical size,” σ/ρ where σ is the
surface resistivity of passive iron surface (in Ω-cm²) and ρ is the
volume resistivity of the acid (in Ω-cm). A simple and informative
structure which demonstrates the essential conditions for strong
electrical coupling between isolated elements of very small size may
be constructed as shown in Figure 4. The dielectric barrier insures
that charge transfer through one dipole must be accompanied by an equal
and opposite transfer through the surfaces of the other dipole. If the
“inexcitable” silver tails have sufficiently high conductance (_i.e._,
sufficiently large surface area, hence preferably, dendrites), strong
coupling will occur, just as though the cores of the two pieces of iron
were connected with a solid conducting wire.
[Illustration: Figure 4]
[Illustration: Figure 5—Electrochemical excitatory-inhibitory
interaction cell]
4. Inhibitory Coupling
If a third “dipole” is inserted through the dielectric membrane in
the opposite direction, then excitation of this isolated element
tends to inhibit the response which would otherwise be elicited by
excitation of one of the parallel dipoles. Figure 5 shows the first
such “logically-complete” interaction cell successfully constructed and
demonstrated. It may be said to behave as an elementary McCulloch-Pitts
neuron (15). Further analysis shows that similar structures
incorporating many dipoles (both excitatory and inhibitory) can be made
to behave as general “linear decision functions” in which all input
weights are approximately proportional to the total size or length of
their corresponding attached dendritic structures.
5. Dendrite Growth
Figure 6 shows a sample gold dendrite grown by electrodeposition
(actual size, about 1 mm) from a 54% nitric acid solution to which gold
chloride was added. When such a dendrite is attached to a piece of
iron (both submerged), activation of the excitable element produces a
field in such a direction as to promote further growth of the dendritic
structure. Thus, if gold chloride is added to the solution used in
the elementary interaction cells described above, all input influence
“weights” tend to increase with use and, hence, produce a plasticity of
function.
6. Field Effects in Locally-Refractory Regions
Our measurements indicate that, during the refractory period following
excitation, the surface resistance of iron in nitric acid drops to
substantially less than 1% of its resting value in a manner reminiscent
of nerve membranes (4). Thus, if a distributed or gross field exists
at any time throughout a complex cellular aggregate, concomitant
current densities in locally-refractive regions will be substantially
higher than elsewhere and, if conditions appropriate to dendrite
growth exist (as described above) growth rates in such regions will
also be substantially higher than elsewhere. It would appear that, as
a result, recently active functional couplings (in contrast to those
not associated with recent neural activity) should be significantly
altered by widely distributed fields or massive peripheral shocks. This
mechanism might thus explain the apparent ability of the brain to form
specific temporal associations in response to spatially-diffuse effects
such as are generated, for example, by the pain receptors.
[Illustration: (a)]
[Illustration: (b)
Figure 6—Dendritic structures, living and non-living. (a) Cat dendrite
trees (from Bok, “Histonomy of the Cerebral Cortex,” Elsevier, 1959);
(b) Electrodeposited gold dendrite tree.]
SUMMARY
An attempt is being made to develop meaningful electrochemical model
techniques which may contribute toward a clearer understanding of
cortical function. Two basic phenomena are simultaneously employed
which are variants of (1) the Lillie iron-wire nerve model, and (2)
growth of metallic dendrites by electrodeposition. These phenomena are
being induced particularly within dense cellular aggregates of various
materials whose interstitial spaces are flooded with liquid electrolyte.
REFERENCES
1. Bok, S. T.,
“Histonomy of the Cerebral Cortex,”
Amsterdam, London:Elsevier Publishing Co., New York:Princeton,
1959
2. Bonhoeffer, K. F.,
“Activation of Passive Iron as a Model for the Excitation of
Nerve,”
_J. Gen. Physiol._ =32=:69-91 (1948).
This paper summarizes work carried out during 1941-1946
at the University of Leipzig, and published during the
war years in German periodicals.
3. Boycott, B. B., and Young, J. Z.,
“The Comparative Study of Learning,”
S. E. B. Symposia, No. IV
“Physiological Mechanisms in Animal Behavior,”
Cambridge: University Press, USA:Academic Press, Inc., 1950
4. Cole, K. S., and Curtis, H. J.,
“Electric Impedance of the Squid Giant Axon During Activity,”
_J. Gen. Physiol._ =22=:649-670 (1939)
5. Eccles, J. C.,
“The Effects of Use and Disuse of Synaptic Function,”
“Brain Mechanisms and Learning—A Symposium,”
organized by the Council for International Organizations of
Medical Science, Oxford:Blackwell Scientific Publications, 1961
6. Franck, U. F.,
“Models for Biological Excitation Processes,”
“Progress in Biophysics and Biophysical Chemistry,”
J. A. V. Butler, ed., London and New York:Pergamon Press,
pp. 171-206, 1956
7. Gerard, R. W.,
“Biological Roots of Psychiatry,”
_Science_ =122 (No. 3162)=:225-230 (1955)
8. Gesell, R.,
“A Neurophysiological Interpretation of the Respiratory Act,”
_Ergedn. Physiol._ =43:=477-639 (1940)
9. Hebb, D. O.,
“The Organization of Behavior, A Neuropsychological Theory,”
New York:John Wiley and Sons, 1949
10. Hebb, D. O.,
“Distinctive Features of Learning in the Higher Animal,”
“Brain Mechanisms and Learning—A Symposium,”
organized by the Council for International Organizations of
Medical Science, Oxford:Blackwell Scientific Publications, 1961
11. Konorski, J.,
“Conditioned Reflexes and Neuron Organization,”
Cambridge:Cambridge University Press, 1948
12. Lillie, R. S.,
“Factors Affecting the Transmission and Recovery in the Passive
Iron Nerve Model,”
_J. Gen. Physiol._ =4=:473 (1925)
13. Lillie, R. S., _Biol. Rev._ =16=:216 (1936)
14. Matumoto, M., and Goto, K.,
“A New Type of Nerve Conduction Model,”
_The Gurma Journal of Medical Sciences_ =4(No. 1)= (1955)
15. McCulloch, W. S., and Pitts, W.,
“A Logical Calculus of the Ideas Immanent in Nervous Activity,”
_Bulletin of Mathematical Biophysics_ =5=:115-133 (1943)
16. Morrell, F.,
“Electrophysiological Contributions to the Neural
Basis of Learning,”
_Physiological Reviews_ =41(No. 3)= (1961)
17. Pask, G.,
“The Growth Process Inside the Cybernetic Machine,”
_Proc. 2nd Congress International Association Cybernetics_,
Gauthier-Villars, Paris:Namur, 1958
18. Retzlaff, E.,
“Neurohistological Basis for the Functioning of Paired
Half-Centers,”
_J. Comp. Neurology_ =101=:407-443 (1954)
19. Sperry, R. W.,
“Neurology and the Mind-Brain Problem,”
_Amer. Scientist_ =40(No. 2)=: 291-312 (1952)
20. Tasaki, I., and Bak, A. F.,
_J. Gen. Physiol._ =42=:899 (1959)
21. Thorpe, W. H.,
“The Concepts of Learning and Their Relation to Those of
Instinct,”
S. E. B. Symposia, No. IV,
“Physiological Mechanisms in Animal Behavior,”
Cambridge:University Press, USA:Academic Press, Inc., 1950
22. Yamagiwa, K.,
“The Interaction in Various Manifestations
(Observations on Lillie’s Nerve Model),”
_Jap. J. Physiol._ =1=:40-54 (1950)
23. Young, J. Z.,
“The Evolution of the Nervous System and of the
Relationship of Organism and Environment,”
G. R. de Beer, ed.,
“Evolution,”
Oxford:Clarendon Press, pp. 179-204, 1938
24. Young, J. Z.,
“Doubt and Certainty in Science, A Biologist’s Reflections
on the Brain,”
New York:Oxford Press, 1951
Multi-Layer Learning Networks
R. A. STAFFORD
_Philco Corp., Aeronutronic Division
Newport Beach, California_
INTRODUCTION
This paper is concerned with the problem of designing a network of
linear threshold elements capable of efficiently adapting its various
sets of weights so as to produce a prescribed input-output relation.
It is to accomplish this adaptation by being repetitively presented
with the various inputs along with the corresponding desired outputs.
We will not be concerned here with the further requirement of various
kinds of ability to “generalize”—_i.e._, to tend to give correct
outputs for inputs that have not previously occurred when they are
similar in some transformed sense to other inputs that have occurred.
In putting forth a model for such an adapting or “learning” network, a
requirement is laid down that the complexity of the adaption process
in terms of interconnections among elements needed for producing
appropriate weight changes, should not greatly exceed that already
required to produce outputs from inputs with a static set of weights.
In fact, it has been found possible to use the output-from-input
computing capacity of the network to help choose proper weight changes
by observing the effect on the output of a variety of possible weight
changes.
No attempt is made here to defend the proposed network model on
theoretical grounds since no effective theory is known at present.
Instead, the plausibility of the various aspects of the network model,
combined with empirical results must suffice.
SINGLE ELEMENTS
To simplify the problem it is assumed that the network receives a set
of two-valued inputs, x₁, x₂, ..., xₙ, and is required to produce only
a single two-valued output, y. It is convenient to assign the numerical
quantities +1 and -1 to the two values of each variable.
The simplest network would consist of a single linear threshold
element with a set of weights, c₀, c₁, c₂, ..., cₙ. These determine
the output-input relation or function so that y is +1 or -1 according
as the quantity, c₀ + c₁x₁ + c₂x₂ + ... + cₙxₙ, is positive or not,
respectively. It is possible for such a single element to exhibit an
adaptive behavior as follows. If, for a given set, x₁, x₂, ..., xₙ, the
output, y, is correct, then make no changes to the weights. Otherwise
change the weights according to the equations
Δc₀ = y* Δcᵢ = y*xᵢ, i = 1,2, ...,n
where y* is the desired output.
It has been shown by a number of people that the weights of such an
element are assured of arriving at a set of values which produce the
correct output-input relation after a sufficient number of errors,
provided that such a set exists. An upper bound on the number of
possible errors can be given which depends only on the initial weight
values and the logical function to be learned. This does not, however,
solve our network problem for two reasons.
First, as the number, n, of inputs gets large, the number of errors
to be expected for most functions which can be learned increases to
unreasonable values. For example, for n = 6, most such functions
result in 500 to 1000 errors compared to an average of 32 errors to be
expected in a perfect learning device.
Second, and more important, the fraction of those logical functions
which can be generated in a single element becomes vanishingly small as
n increases. For example, at n = 6 less than one in each three trillion
logical functions is so obtainable.
NETWORKS OF ELEMENTS
It can be demonstrated that if a sufficiently large number of linear
threshold elements is used, with the outputs of some being the inputs
of others, then a final output can be produced which is any desired
logical function of the inputs. The difficulty in such a network lies
in the fact that we are no longer provided with a knowledge of the
correct output for each element, but only for the final output. If the
final output is incorrect there is no obvious way to determine which
sets of weights should be altered.
As a result of considerable study and experimentation at Aeronutronic,
a network model has been evolved which, it is felt, will get around
these difficulties. It consists of four basic features which will now
be described.
Positive Interconnecting Weights
It is proposed that all weights in elements attached to inputs which
come from other elements in the network be restricted to positive
values. (Weights attached to the original inputs to the network, of
course, must be allowed to be of either sign.) The reason for such a
restriction is this. If element 1 is an input to element 2 with weight
c₁₂, element 2 to element 3 with weight c₂₃, _etc._, then the sign of
the product, c₁₂c₂₃ ..., gives the sense of the effect of a change in
the output of element 1 on the final element in the chain (assuming
this is the only such chain between the two elements). If these various
weights were of either possible sign, then a decision as to whether or
not to change the output in element 1 to help correct an error in the
final element would involve all weights in the chain. Moreover, since
there would in general be a multiplicity of such chains, the decision
is rendered impossibly difficult.
The above restriction removes this difficulty. If the output of any
element in the network is changed, say, from -1 to +1, the effect on
the final element, if it is affected at all, is in the same direction.
It should be noted that this restriction does not seriously affect
the logical capabilities of a network. In fact, if a certain logical
function can be achieved in a network with the use of weights of
unrestricted sign, then the same function can be generated in another
network with only positive interconnecting weights and, at worst, twice
the number of elements. In the worst case this is done by generating
in the restricted network both the output and its complement for each
element of the unrestricted network. (It is assumed that there are no
loops in the network.)
A Variable Bias
The central problem in network learning is that of determining, for
a given input, the set of elements whose outputs can be altered so
as to correct the final element, and which will do the least amount
of damage to previous adaptations to other inputs. Once this set has
been determined, the incrementing rule given for a single element will
apply in this case as well (subject to the restriction of leaving
interconnecting weights positive), since the desired final output
coincides with that desired for each of the elements to be changed
(because of positive interconnecting weights).
In the process of arriving at such a decision three factors need to be
considered. Elements selected for change should tend to be those whose
output would thereby be affected for a minimum number of other possible
inputs. At the same time it should be ascertained that a change in
each of the elements in question does indeed contribute significantly
towards correcting the final output. Finally, a minimum number of such
elements should be used.
It would appear at first that this kind of decision is impossible to
achieve if the complexity of the decision apparatus is kept comparable
to that of the basic input-output network as mentioned earlier.
However, in the method to be described it is felt that a reasonable
approximation to these requirements will be achieved without an undue
increase in complexity.
It is assumed that in addition to its normal inputs, each element
receives a variable input bias which we can call b. The output of every
element should then be determined by the sign of the usual weighted
sum of its inputs plus this bias quantity. This bias is to be the same
for each element of the network. If b = 0 the network will behave
as before. However, if b is increased gradually, various elements
throughout the network will commence changing from -1 to +1, with one
or a few changing at any one time as a rule. If b is decreased, the
opposite will occur.
Now suppose that for a given input the final output ought to be +1 but
actually is -1. Assume that b is then raised so high that this final
output is corrected. Then commence a gradual decline in b. Various
elements may revert to -1, but until the final output does, no weights
are changed. When the final output does revert to -1, it is due to an
element’s having a sum (weighted sum plus bias) which just passed down
through zero. This then caused a chain effect of changing elements
up to the final element, but presumably this element is the only one
possessing a zero sum. This can then be the signal for the weights
on an element to change—a change of final output from right to wrong
accompanied simultaneously by a zero sum in the element itself.
After such a weight change, the final output will be correct once more
and the bias can again proceed to fall. Before it reaches zero, this
process may occur a number of times throughout the network. When the
bias finally stands at zero with the final output correct, the network
is ready for the next input. Of course if -1 is desired, the bias will
change in the opposite direction.
It is possible that extending the weight change process a little past
the zero bias level may have beneficial results. This might increase
the life expectancy of each learned input-output combination and
thereby reduce the total number of errors. This is because the method
used above can stop the weight correction process so that even though
the final output is correct, some elements whose output are essential
to the final output have sums close to zero, which are easily changed
by subsequent weight changes.
It will be noted that this method conforms to all three considerations
mentioned previously. First, by furnishing each element the same bias,
and by not changing weights until the final output becomes incorrect
with dropping bias, there is a strong tendency to select elements
which, with b = 0, would have sums close to zero. But the size of the
sum in an element is a good measure of the amount of damage done to
an element for other inputs if its current output is to be changed.
Second, it is obvious that each element changed has had a demonstrable
effect on the final output. Finally, there will be a clear tendency to
change only a minimum of elements because changes never occur until the
output clearly requires a change.
On the other hand this method requires little more added complexity to
the network than it already has. Each element requires a bias, an error
signal, and the desired final output, these things being uniform for
all elements in a network. Some external device must manipulate the
bias properly, but this is a simple behavior depending only on an error
signal and the desired final output—not on the state of individual
elements in the network. What one has, then, is a network consisting
of elements which are nearly autonomous as regards their decisions
to change weights. Such a scheme appears to be the only way to avoid
constructing a central weight-change decision apparatus of great
complexity. This rather sophisticated decision is made possible by
utilizing the computational capabilities the network already possesses
in producing outputs from inputs.
It should be noted here that this varying bias method requires that
the variable bias be furnished to just those elements which have
variable weights and to no others. Any fixed portion of the network,
such as preliminary layers or final majority function for example,
must operate independently of the variable bias. Otherwise, the final
output may go from right to wrong as the bias moves towards zero and no
variable-weight element be to blame. In such a case the network would
be hung up.
Logical Redundancy in the Network
A third aspect of the network model is that for all the care taken
in the previous steps, they will not suffice in settling quickly to
a set of weights that will generate the required logical function
unless there is a great multiplicity of ways in which this can be done.
This is to say that a learning network needs to have an excess margin
of weights and elements beyond the minimum required to generate the
functions which are to be learned.
This is analogous to the situation that prevails for a single element
as regards the allowed range of values on its weights. It can be shown
for example, that any function for n=6 that can be generated by a
single element can be obtained with each weight restricted to the range
of integer values -9,-8, ..., +9. Yet no modification of the stated
weight change rule is known which restricts weight values to these and
yet has any chance of ever being learned for most functions.
Fatigued Elements
It would appear from some of the preliminary results of network
simulations that it may be useful to have elements become “fatigued”
after undergoing an excessive number of weight changes. Experiments
have been performed on simplifications of the model described so far
which had the occasional result that a small number of elements came
to a state where they received most of the weight increments, much
to the detriment of the learning process. In such cases the network
behaves as if it were composed of many fewer adjustable elements. In a
sense this is asking each element to maintain a record of the data it
is being asked to store so that it does not attempt to exceed its own
information capacity.
It is not certain just how this fatigue factor should enter in the
element’s actions, but if it is to be compatible with the variable bias
method, this fatigue factor must enter into the element’s response to
a changing bias. Once an element changes state with zero sum at the
same time that the final output becomes wrong, incrementing must occur
if the method is to work. Hence a “fatigued” element must respond less
energetically to a change of bias, perhaps with a kind of variable
factor to be multiplied by the bias term.
NETWORK STRUCTURE
It is felt that the problem of selecting the structure of
interconnections for a network is intimately connected to the
previously mentioned problem of generalization. Presumably a given
type of generalization can be obtained by providing appropriate fixed
portions of the network and an appropriate interconnection structure
for the variable portion. However, for very large networks, it is
undoubtedly necessary to restrict the complexity so that it can be
specified by relatively simple rules. Since very little is known about
this quite important problem, no further discussion will be attempted
here.
COMPUTER SIMULATION RESULTS
A computer simulation of some of the network features previously
described has been made on an IBM 7090. Networks with an excess of
elements and with only positive interconnecting weights were used.
However, in place of the variable bias method, a simple choice of the
element of sum closest to, and on the wrong side of, zero was made
without regard to the effectiveness of the element in correcting the
final output. No fatigue factors were used.
The results of these simulations are very encouraging, but at the same
time indicate the need for the more sophisticated methods. No attempt
will be made here to describe the results completely.
In one series of learning experiments, a 22-element network was used
which had three layers, 10 elements on the first, 11 on the second, and
1 on the third. The single element on the third was the final output,
and was a fixed majority function of the 11 elements in the second
layer. These in turn each received inputs from each of the 10 on the
first layer and from each of the 6 basic inputs. The 10 on the first
layer each received only the 6 basic inputs. A set of four logical
functions, A, B, C, and D, was used. Function A was actually a linear
threshold function which could be generated by the weights 8, 7, 6, 5,
4, 3, 2, functions B and C were chosen by randomly filling in a truth
table, while D was the parity function.
TABLE I
-------+--------+---------+-------
A | B | C | D
r e | r e | r e | r e
-------+--------+---------+-------
5 54 | 8 100 | 11 101 | 4 52
4 37 | 9 85 | 4 60 | 5 62
4 44 | 6 72 | 9 85 | 6 56
-------+--------+---------+-------
Table I gives the results of one series of runs with these functions
and this network, starting with various random initial weights. The
quantity, r, is the number of complete passes through the 64-entry
truth table before the function was completely learned, while e is
the total number of errors made. In evaluating the results it should
be noted that an ideal learning device would make an average of 32
errors altogether on each run. The totals recorded in these runs are
agreeably close to this ideal. As expected, the linear threshold
function is the easiest to learn, but it is surprising that the
parity function was substantially easier than the two randomly chosen
functions. Table II gives a chastening result of the same experiment
with all interconnecting weights removed except that the final element
is a fixed majority function of the other 21 elements. Thus there was
adaptation on one layer only. As can be seen Table I is hardly better
than Table II so that the value of variable interconnecting weights was
not being fully realized. In a later experiment the number of elements
was reduced to 12 elements and the same functions used. In this case
the presence of extra interconnecting weights actually proved to be
a hindrance! However a close examination of the incrementing process
brought out the fact that the troublesome behavior was due to the
greater chance of having only a few (often only one) elements do nearly
all the incrementing. It is expected that the use of the additional
refinements discussed herein will produce a considerable improvement
in bringing out the full power of adaptation in multiple layers of a
network.
TABLE II
-------+--------+---------+-------
A | B | C | D
r e | r e | r e | r e
-------+--------+---------+-------
7 47 | 18 192 | 8 110 | 4 48
3 40 | 7 69 | 10 98 | 6 68
4 43 | 7 82 | 4 47 | 6 46
-------+--------+---------+-------
FUTURE PROBLEMS
Aside from the previous question of deciding on network structure,
there are several other questions that remain to be studied in learning
networks.
There is the question of requiring more than a single output from a
network. If, say, two outputs are required for a given input, one
+1 and the other -1, this runs into conflict with the incrementing
process. Changes that aid one output may act against the other.
Apparently the searching process depicted before with a varying bias
must be considerably refined to find weight changes which act on
all the outputs in the required way. This is far from an academic
question because there will undoubtedly be numerous cases in which
the greatest part of the input-output computation will have shared
features for all output variables. Only at later levels do they need to
be differentiated. Hence it is necessary to envision a single network
producing multiple outputs rather than a separate network for each
output variable if full efficiency is to be achieved.
Another related question is that of using input variables that are
either many-, or continuous-, valued rather than two-valued. No
fundamental difficulties are discernible in this case, but the matter
deserves some considerable study and experimentation.
Another important question involves the use of a succession of inputs
for producing an output. That is, it may be useful to allow time to
enter into the network’s logical action, thus giving it a “dynamic” as
well as “static” capability.
Adaptive Detection of Unknown Binary Waveforms
J. J. SPILKER, JR.
_Philco Western Development Laboratories
Palo Alto, California_
This work was supported by the Philco WDL
Independent Development Program. This paper,
submitted after the Symposium, represents a
more detailed presentation of some of the
issues raised in the discussion sessions at the
Symposium and hence, constitutes a worthwhile
addition to the Proceedings.
INTRODUCTION
One of the most important objectives in processing a stream of
data is to determine and detect the presence of any invariant or
quasi-invariant “features” in that data stream. These features are
often initially unknown and must be “learned” from the observations.
One of the simplest features of this form is a finite length signal
which occurs repetitively, but not necessarily periodically with time,
and has a waveshape that remains invariant or varies only slowly with
time.
In this discussion, we assume that the data stream has been
pre-processed, perhaps by a detector or discriminator, so as to exhibit
this type of repetitive (but unknown) waveshape or signal structure.
The observed signal, however, is perturbed by additive noise or other
disturbances. It is desired to separate the quasi-invariance of the
data from the truly random environment. The repetitive waveform may
represent, for example, the transmission of an unknown sonar or radar,
a pulse-position modulated noise-like waveform, or a repeated code word.
The problem of concern is to estimate the signal waveshape and to
determine the time of each signal occurrence. We limit this discussion
to the situation where only a single repetitive waveform is present
and the signal sample values are binary. The observed waveform is
assumed to be received at low signal-to-noise ratio so that a single
observation of the signal (even if one knew precisely the arrival time)
is not sufficient to provide a good estimate of the signal waveshape.
The occurrence time of each signal is assumed to be random.
THE ADAPTIVE DETECTION MACHINE
The purpose of this note is to describe very briefly a machine[2] which
has been implemented to recover the noise-perturbed binary waveform.
A simplified block diagram of the machine is shown in Figure 1. The
experimental machine has been designed to operate on signals of 10³
samples duration.
[2] The operation of this machine is described in substantially greater
detail in J. J. Spilker, Jr., D. D. Luby, R. D. Lawhorn, “Adaptive
Binary Waveform Detection,” Philco Western Development Laboratories,
Communication Sciences Department, TR #75, December 1963.
Each analog input sample enters the machine at left and may either
contain a signal sample plus noise or noise alone. In order to permit
digital operation in the machine, the samples are quantized in a
symmetrical three-level quantizer. The samples are then converted
to vector form, _e.g._, the previous 10³ samples form the vector
components. A new input vector, ⮕Y⁽ⁱ⁾, is formed at each sample instant.
Define the signal sample values as s₁, s₂, ..., sₙ. The observed vector
Y⁽ⁱ⁾ is then either (a) perfectly centered signal plus noise, (b)
shifted signal plus noise, or (c) noise alone.
{ (s₁, s₂, ..., sₙ) + (n₁, n₂, ..., nₙ) (a)
(Y⁽ⁱ⁾)ᵗ = { (0, ..., s₁, s₂, ..., sₙ₋ⱼ) + (n₁, n₂, ..., nₙ) (b)
{ (0 ... 0) + (n₁, n₂, ..., nₙ) (c)
At each sample instant, two measurements are made on the input
vector, an energy measurement ‖Y⁽ⁱ⁾‖² and a polarity coincidence
cross-correlation with the present estimate of the signal vector stored
in memory. If the weighted sum of the energy and cross-correlation
measurements exceeds the present threshold value Γᵢ, the input vector
is accepted as containing the signal (properly shifted in time), and
the input vector is added to the memory. The adaptive memory has 2^{Q}
levels, 2^{Q-1} positive levels, 1 zero level and 2^{Q-1}-1 negative
levels. New contributions are made to the memory by normal vector
addition except that saturation occurs when a component value is at the
maximum or minimum level.
The acceptance or rejection of a given input vector is based on a
hypersphere decision boundary. The input vector is accepted if the
weighted sum γᵢ exceeds the threshold Γᵢ
γᵢ = Y⁽ⁱ⁾∙M⁽ⁱ⁾ + α‖Y⁽ⁱ⁾‖² ⩾ Γᵢ.
[Illustration: Figure 1—Block diagram of the adaptive binary waveform
detector]
Geometrically, we see that the input vector is accepted if it falls on
or outside of a hypersphere centered at ⮕C⁽ⁱ⁾ = -⮕M⁽ⁱ⁾/2α having radius
squared
Γ⁽ⁱ⁾ ‖M⁽ⁱ⁾‖²
[r⁽ⁱ⁾]² = ——— + —————— .
α (2α)²
Both the center and radius of this hypersphere change as the machine
adapts. The performance and optimality of hypersphere-type decision
boundaries have been _discussed in related work_ by Glaser[3] and
Cooper.[4]
[3] F. M. Glaser, “Signal Detection by Adaptive Filters,” _IRE Trans.
Information Theory_, pp. 87-90; April 1961.
[4] P. W. Cooper, “The Hypersphere in Pattern Recognition,”
_Information and Control_, pp. 324-346; December 1962.
The threshold value, Γᵢ, is adapted so that it increases if the
memory becomes a better replica of the signal with the result that γᵢ
increases. On the other hand, if the memory is a poor replica of the
signal (for example, if it contains noise alone), it is necessary that
the threshold decay with time to the point where additional acceptances
can modify the memory structure.
The experimental machine is entirely digital in operation and, as
stated above, is capable of recovering waveforms of up to 10³ samples
in duration. In a typical experiment, one might attempt to recover
an unknown noise-perturbed, pseudo-random waveform of up to 10³ bits
duration which occurs at random intervals. If no information is
available as to the signal waveshape, the adaptive memory is blank at
the start of the experiment.
In order to illustrate the operation of the machine most clearly, let
us consider a repetitive binary waveform which is composed of 10³ bits
of alternate “zeros” and “ones.” A portion of this waveform is shown in
Figure 2a. The waveform actually observed is a noise-perturbed version
of this waveform shown in Figure 2b at-6 db signal-to-noise ratio. The
exact sign of each of the signal bits obviously could not be accurately
determined by direct observation of Figure 2b.
[Illustration: (a) Binary signal]
[Illustration: (b) Binary signal plus noise
Figure 2—Binary signal with additive noise at-6 db SNR]
[Illustration: (a) (b)]
[Illustration: (c) (d)]
[Illustration: (e)
Figure 3—Adaption of the memory at-6 db SNR: (a) Blank initial memory;
(b) Memory after first dump; (c) Memory after 12 dumps; (d) Memory
after 40 dumps; (e) Perfect “checkerboard” memory for comparison]
As the machine memory adapts to this noisy input signal, it progresses
as shown in Figure 3. The sign of 10^{3} memory components are
displayed in a raster pattern in this figure. Figure 3a shows the
memory in its blank initial state at the start of the adaption process.
Figure 3b shows the memory after the first adaption of the memory. This
first “dump” occurred after the threshold had decayed to the point
where an energy measurement produced an acceptance decision. Figure
3c and 3d show the memory after 12 and 40 adaptions, respectively.
These dumps, of course, are based on both energy and cross-correlation
measurements. As can be seen, the adapted memory after 40 dumps is
already quite close to the perfect memory shown by the “checkerboard”
pattern of Figure 3c.
The detailed analysis of the performance of this type of machine
vs. signal-to-noise ratio, average signal repetition rate, signal
duration, and machine parameters is extremely complex. Therefore, it
is not appropriate here to detail the results of the analytical and
experimental work on the performance of this machine. However, several
conclusions of a general nature can be stated.
(a) Because the machine memory is always adapting, there
is a relatively high penalty for “false alarms.”
False alarms can destroy a perfect memory. Hence,
the threshold level needs to be set appropriately
high for the memory adaption. If one wishes to
detect signal occurrences with more tolerance to
false alarms, a separate comparator and threshold
level should be used.
(b) The present machine structure, which allows for
slowly varying changes in the signal waveshape,
exhibits a marked threshold effect in steady-state
performance at an input signal-to-noise ratio
(peak signal power-to-average noise power ratio)
of about -12 db. Below this signal level, the time
required for convergence increases very rapidly with
decreasing signal level. At higher SNR, convergence
to noise-like signals, having good auto-correlation
properties, occurs at a satisfactory rate.
A more detailed discussion of performance has been published in the
report cited in footnote reference 1.
Conceptual Design of Self-Organizing Machines
P. A. KLEYN
_Northrop Nortronics_
_Systems Support Department_
_Anaheim, California_
Self-organization is defined and several examples
which motivate this definition are presented. The
significance of this definition is explored by
comparison with the metrization problem discussed
in the companion paper (1) and it is seen that
self-organization requires decomposing the space
representing the environment. In the absence
of a priori knowledge of the environment, the
self-organizing machine must resort to a sequence
of projections on unit spheres to effect this
decomposition. Such a sequence of projections
can be provided by repeated use of a nilpotent
projection operator (NPO). An analog computer
mechanization of one such NPO is discussed
and the signal processing behavior of the NPO
is presented in detail using the Euclidean
geometrical representation of the metrizable
topology provided in the companion paper.
Self-organizing systems using multiple NPO’s
are discussed and current areas of research are
identified.
INTRODUCTION
Unlike the companion paper which considers certain questions in
depth, this paper presents a survey of the scope of our work in
self-organizing systems and is not intended to be profound.
The approach we have followed may be called phenomenological (Figure
1). That is, the desired behavior (self-organization) was defined,
represented mathematically, and a mechanism(s) required to yield the
postulated behavior was synthesized using mathematical techniques. One
advantage of this approach is that it avoids assumptions of uniqueness
of the mechanism. Another advantage is that the desired behavior, which
is after all the principal objective, is taken as invariant. An obvious
disadvantage is the requirement for the aforementioned synthesis
technique; fortunately in our case a sufficiently general technique had
been developed by the author of the companion paper.
From the foregoing and from the definition of self-organization we
employ (see conceptual model), it would appear that our research does
not fit comfortably within any of the well publicized approaches to
self-organization (2). Philosophically, we lean toward viewpoints
expressed by Ashby (3), (4), Hawkins (5), and Mesarovic (6) but with
certain reservations. We have avoided the neural net approach partly
because it is receiving considerable attention and also because the
brain mechanism need not be the unique way to produce the desired
behavior.
[Illustration: Figure 1—Approach used in Nortronics research on
self-organizing systems]
Nor have we followed the probability computer or statistical decision
theory approach exemplified by Braverman (7) because these usually
require some sort of preassigned coordinate system (8). Neither will
the reader find much indication of formal logic (9) or heuristic (10)
programming. Instead, we view a self-organizing system more as a mirror
whose appearance reflects the environment rather than its own intrinsic
nature. With this viewpoint, a self-organizing system appears very
flexible because it possesses few internal constraints which would tend
to distort the reflection of the environment and hinder its ability to
adapt.
CONCEPTUAL MODEL
Definition
A system is said to be self-organizing if, after observing the input
and output of an unknown phenomenon (transfer relation), the system
organizes itself into a simulation of the unknown phenomenon.
Implicit in this definition is the requirement that the self-organizing
machine (SOM) not possess a preassigned coordinate system. In fact it
is just this ability to acquire that coordinate system implicit in the
input-output spaces which define the phenomenon that we designate as
self-organization. Thus any a priori information programmed into the
SOM by means of, for example, stored or wired programs, constrains
the SOM and limits its ability to adapt. We do not mean to suggest
that such preprogramming is not useful or desirable; merely that it is
inconsistent with the requirement for self-organization. As shown in
Figure 2, it is the given portion of the environment which the SOM is
to simulate, which via the defining end spaces, furnishes the SOM with
all the data it needs to construct the coordinate system intrinsic to
those spaces.
The motivation for requiring the ability to simulate as a feature of
self-organization stems from the following examples.
Consider the operation of driving an automobile. Figure 3 depicts the
relation characterized by a set of inputs; steering, throttle, brakes,
transmission, and a set of outputs; the trajectory. Operation of the
automobile requires a device (SOM) which for a desired trajectory can
furnish those inputs which realize the desired trajectory. In order to
provide the proper inputs to the automobile, the SOM must contain a
simulation of ⨍⁻¹(x).
[Illustration: Figure 2—Simulation of (a portion of) the environment]
[Illustration: Figure 3—Simulation of a relation]
Since ⨍(x) is completely defined in terms of the inputs and the
resulting trajectories, exposure to them provide the SOM with all the
information necessary to simulate ⨍⁻¹(x). And if the SOM possesses
internal processes which cause rearrangement of the input-output
relation of the SOM to correspond to ⨍⁻¹(x) in accordance with the
observed data, the SOM can operate an automobile. It is this internal
change which is implied by the term “self-organizing,” but note that
the instructions which specify the desired organization have their
source in the environment.
As a second example consider adaptation to the environment. Adapt
(from Webster) means: “to change (oneself) so that one’s behavior,
attitudes, _etc._, will conform to new or changed circumstances.
Adaptation in biology means a change in structure, function or form
that produces better adjustment to the environment.” These statements
suggest a simulation because adjustment to the environment implies
survival by exposing the organism to the beneficial rather than the
inimical effects of the environment. If we represent the environment
(or portion thereof) as a relation as shown in Figure 2, we note that
the ability to predict what effect a given disturbance will have is due
to a simulation of the cause-effect relation which characterizes the
environment.
It would be a mistake to infer from these examples that simulation
preserves the appearance of the causes and effects which characterize
a relation. We clarify this situation by examining a relation and its
simulation.
Consider the relation between two mothers and their sons as pictured
in Figure 4. Observe that if symbols (points) are substituted for the
actual physical objects (mothers and sons), the relation is not altered
in any way. This is what we mean by simulation and this is how a SOM
simulates. It is not even necessary that the objects, used to display
the relation, be defined; _i.e._, these objects may be primitive.
(If this were not so, no mathematical or physical theory could model
the environment.) The main prerequisite is sufficient resolution to
distinguish the objects from each other.
[Illustration: Figure 4—A relation of objects—displayed and simulated]
MATHEMATICAL MODEL
The mathematical model must represent both the environment and the SOM
and for reasons given in the companion paper each is represented as a
metrizable topology. For uniqueness we factor each space into equal
parts and represent the environment as the channel
W ⟶ X. (Ref. 10a)
Consider now the SOM to be represented by the cascaded channels
X ⟶ Y ⟶ Z
where X ⟶ Y is a variable which represents the reorganization of the
SOM existing input-output relation represented by Y ⟶ Z.
The solution of the three channels-in-cascade problem
W ⟶ X ⟶ Y ⟶ Z,
where p(W) (11), p(X), p(X|W), p(Y), p(Z), p(Z|Y) are fixed, yields
that middle channel p₀(Y|X), from a set of permissible middle channels
{p(Y|X)}, which maximizes R(Z,W).
Then the resulting middle channel describes that reorganization of the
SOM which yields the optimum simulation of W ⟶ X by the SOM, within the
constraints upon Ch(Z,Y).
The solution (the middle channel) depends of course on the particular
end channels. Obviously the algorithm which is used to find the
solution does not. It follows that if some physical process were
constrained to carrying out the steps specified by the algorithm,
said process would be capable of simulation and would exhibit
self-organization.
Although the formal solution to the three-channels-in-cascade problem
is not complete, the solution is sufficiently well characterized to
permit proceeding with a mechanization of the algorithm. A considerable
portion of the solution is concerned with the decomposition and
metrization of channels and it is upon this feature that we now focus
attention.
As suggested in the companion paper, if the dimensionality of the
spaces is greater than one, the SOM has only one method available (12).
Consider the decomposition of a space without, for the moment, making
the distinction between input and output.
Figure 5 depicts objects represented by a (perhaps multidimensional)
“cloud” of points. In the absence of a preassigned coordinate system,
the SOM computes the center of gravity of the cloud (which can be
done in any coordinate system) and describes the points in terms of
the distance from this center of gravity; or, which is the same, as
concentric spheres with origin at the center of gravity.
[Illustration: Figure 5—Nilpotent decomposition of a three-dimensional
space]
The direction of particular point cannot be specified for there is no
reference radius vector. Since the SOM wants to end up with a cartesian
coordinate system, it must transform the sphere (a two-dimensional
surface) into a plane (a two-dimensional surface). Unfortunately, a
sphere is not homeomorphic to a plane; thus the SOM has to decompose
the sphere into a cartesian product of a hemisphere (12a) and a
denumerable group. The SOM then can transform the hemisphere into a
plane. The points projected onto the plane constitute a space of the
same character as the one with which the SOM started. Thus, it can
repeat all operations on the plane (a space of one less dimension) by
finding the center of gravity and the circle upon which the desired
point is situated. The circle is similarly decomposed into a line times
a denumerable group. By repeating this operation as many times as the
space has dimensions, the SOM eventually arrives at a single point and
has obtained in the process a description of the space. Since this
procedure can be carried on by the repeated use of one operator, this
operator is nilpotent and to reflect this fact as well as the use of a
projection, we have named this a nilpotent projection operator or NPO
for short.
MECHANIZATION OF THE NPO
Analog computer elements were used to simulate one NPO which was
tested in the experimental configuration shown in Figure 6. The NPO
operates upon a channel which is artificially generated from the two
noise generators i₁ and i₂ and the signal generator i₀ (i₀ may also be
a noise generator). The NPO accepts the inputs labelled X₁ and X₂ and
provides the three outputs Ξ₁, Ξ₂, and γ. X₁ is the linear combination
of the outputs of generators i₁ and i₀, similarly X₂ is obtained from
i₂ and i₀.
[Illustration: Figure 6—Experimental test configuration for the
simulation of an NPO]
Obviously, i₀ is an important parameter since it represents the memory
relating the spaces X₁ and X₂. Ξ₁ has the property that the magnitude
of its projection on i₀ is a maximum while Ξ₂ to the opposite has a
zero projection on i₀. γ is the detected version of the eigenvalue of
Ch(X₂,X₁).
In the companion paper it was shown how one can provide a Euclidean
geometrical representation of the NPO. This representation is shown in
Figure 7 which shows the vectors i₀, i₁, i₂, X₁, X₂, Ξ₁, Ξ₂, and the
angles Θ₁, Θ₂, and γ. The length of a vector is given by
|X| = κₓ(2πε)⁻¹ᐟ² ∈ H(X)
and the angle between two vectors by
|Θ(X₁,X₂)|-sin⁻¹ ∈ -R(X₁,X₂).
The three vectors i₀, i₁, i₂ provide an orthogonal coordinate system
because the corresponding signals are random, _i.e._,
κ
R(i₀,i₁,i₂) ≡ 0.
As external observers we have a prior knowledge of this coordinate
system; however, the NPO is given only the vectors X₁ and X₂ in the i₀
⨉ i₁ and i₀ ⨉ i₂ planes. The NPO can reconstruct the entire geometry
but the actual output Ξ obviously is constrained to lie in the plane of
the input vector X. The following formulas are typical of the relations
present.
|Ξ₁|
tan β = ————
|Ξ₂|
cos Θ = cos 2β csc 2γ
cos 2β
cos 2Θ₁ = -1 + 2 ———————
1-cos 2γ
cos Θ = cos Θ₁ cos Θ₂.
[Illustration: Figure 7—Geometry of the NPO]
[Illustration: Figure 8—NPO run number 5]
[Illustration: Figure 9—NPO run number 6]
We have obtained a complete description of the NPO which involves 74
formulas. These treat the noise in the various outputs, invariances of
the NPO and other interesting features. A presentation of these would
be outside of the scope of this paper and would tend to obscure the
main features of the NPO. Thus, we show here only a typical sample of
the computer simulation, Figure 8 and Figure 9. Conditions for these
runs are shown in Table I. Run No. 6 duplicates run No. 5 except for
the fact that i₁ and i₂ were disabled in run No. 6.
Observe that all our descriptions of the NPO and the space it is to
decompose have been time invariant while the signals shown in the
simulation are presented as functions of time. The conversion may be
effected as follows: Given a measurable (single-valued) function
x = x(t)t ∊ T
where
μ(T) > 0
we define the space
X = {x = x(t) ∍ t ∊ T}
and a probability distribution
μ(x⁻¹(X′))
P(X′) = —————————— X′ open ⊂ X
μ(T)
on that space.
TABLE I
Legend for Traces of Figures 8 and 9
---------+-------+-------+--------+-----+----------+-------+--------
Trace | | | | | | |
Number | 1 | 2 | 3 | 4 | 5 | 6 | 7
---------+-------+-------+--------+-----+----------+-------+--------
Symbol | X₂ | X₁ | γ | β | i | dξ₂/dτ | dξ₁/dτ
---------+-------+-------+--------+-----+----------+--------+-------
run No. 5| | | | | | |
| | | | | | |
signal |7½ Vrms|7½ Vrms| π ptop | |35.6 m cps| |
| | | | | | |
noise |16 Vrms|15 Vrms| π/9 | | | |
| | | ptop[5]| |sine wave | |
| | | | | | |
DC | 0 | 0 | | | | |
| | | | | | |
power s/n| 1/4 | 1/4 | 81/1 | | | 0 | 1/2[6]
| | | | | | |
terminal | | | | | | |
value | | | π/4 | π/4 | | |
---------+-------+-------+--------+-----+----------+--------+-------
run No. 6| | | | | | |
| | | | | | |
signal |7½ Vrms|7½ Vrms| π ptop | |35.6 m cps| |
| | | | | | |
noise | 0 | 0 | 0[7] | |sine wave | |
| | | | | | |
DC | -30V | 0 | | | | |
| | | | | | |
power s/n| ∞ | ∞ | ∞ | | | 0 | ∞
| | | | | | |
terminal | | | | | | |
value | | | π/4 | π/4 | | |
---------+-------+-------+--------+-----+----------+--------+-------
[5] Observed from Oscillogram
[6] Computed
[7] Observed from Oscillogram
Then (X,p(X)) is a stochastic space in our usual sense and x(T) is a
stochastic variable. Two immediate consequences are:
P(X) is stationary (P(X) is not a function of t ∊ T), and no question
of ergodicity arises.
NETWORKS OF NPO’S
A network of NPO’s may constitute anything from a SOM to a
preprogrammed detector, depending upon the relative amount of
preprogramming included. Two methods of preprogramming are: (1) Feeding
a signal out of a permanent storage into some of the inputs of the
network of NPO’s. This a priori copy need not be perfect, because the
SOM will measure the angles Θᵢ anyhow. (2) Feedback, which, after all,
is just a way of taking advantage of the storage inherent in any delay
line. (We implicitly assume that any reasonable physical realization
of an NPO will include a delay T between the x input and the ξ output
which is not less than perhaps 10⁻¹ times the time constant of the
internal feedback loop in the γ computation.)
Simulation of channels that possess a discrete component requires
feedback path(s) to generate the required free products of the finitely
generated groups. Then, such a SOM converges to a maximal subgroup of
the group describing the symmetry of the signal that is a free product
available to this SOM.
Because a single NPO with 1 ≤ n₀ ≤ K₀ is isomorphic (provides the same
input to output mapping) to a suitable network of NPO’s with n₀ = 1, it
suffices to study only networks of NPO’s with n₀ = 1.
Figure 10 is largely self-explanatory. Item a is our schematic symbol
for a single NPO with n₀ = 1. Items b, d (including larger feedback
loops), and f are typical of artificial intelligence networks. Item c
is employed to effect the level changing required in order to apply the
three channels in cascade algorithm to the solution of one-dimensional
coding problems. Observe that items c and e are the only configurations
requiring the γ output. Item d may be used as a limiter by making T⁻¹
high compared to the highest frequency present in the signal. Observe
that item e is the only application of NPO’s that requires either the
ξ₂ or β outputs. Item f serves the purpose of handling higher power
levels into and out of what effectively is a single (larger) NPO.
[Illustration]
[Illustration: Figure 10—Some possible networks of NPO’s]
CONCLUSION
The definition of self-organizing behavior suitably represented has
permitted the use of Information Theoretic techniques to synthesize
a (mathematical) mechanism for a self-organizing machine. Physical
mechanization in the form of an NPO has been accomplished and has
introduced the experimental phase of the program. From among the many
items deserving of further study we may mention: more economical
physical mechanization through introduction of modern technology;
identification of networks of NPO’s with their group theoretic
descriptions; analysis of the dimensionality of tasks which a SOM might
be called on to simulate, and prototype SOM applications to related
tasks. It is hoped that progress along these lines can be reported in
the future.
REFERENCES
1. Ścibor-Marchocki, Romuald I.,
“A Topological Foundation for Self-Organization,”
Anaheim, California:Northrop Nortronics, NSS Report 2828,
November 14, 1963
2. It is true that our definition is very similar to that proposed
by Hawkins (reference 5). Compare for example his definition of
learning machines (page 31 of reference 5). But the subsequent
developments reviewed therein are different from the one we have
followed.
3. Ashby, W. R.,
“The Set Theory of Mechanism and Homeostasis,”
Technical Report 7, University of Illinois, September 1962
4. Ashby, W. R.,
“Systems and Information,”
_Transactions PTGME_ =MIL-7=:94-97 (April-July, 1963)
5. Hawkins, J. K.,
“Self-Organizing Systems—A Review and Commentary,”
_Proc. IRE_. =49=:31-48 (January 1961)
6. Mesarovic, M. D.,
“On Self Organizational Systems,”
Spartan Books, pp. 9-36, 1962
7. Braverman, D.,
“Learning Filters for Optimum Pattern Recognition,”
_PGIT_ =IT-8=:280-285 (July 1962)
8. We make the latter statement despite the fact that we employ a
statistical treatment of self-organization. We may predict the
performance of, for example, the NPO by using a statistical
description, but it does not necessarily follow that the NPO
computes statistics.
9. McCulloch, W. S., and Pitts, W.,
“A Logical Calculus of the Ideas Imminent in Nervous Activity,”
_Bull-Math. Biophys_ =5=:115 (1943)
10. Newell, A., Shaw, J. C., and Simon, H. A.,
“Empirical Explorations of the Logic Theory Machine:
A Case Study in Heuristic,”
_Proc. WJCC_, pp. 218-230, 1957
10a. The spaces W, X, Y, and Z are stochastic spaces; that is,
each space is defined as the ordered pair (X,p(X)) where
p(X) = {p(x) ∋ x ∈ X}, p(x) ≥ 0, x ∈ X and ∫x p(x)dx = 1.
Such spaces possess a metrizable topology.
11. We use the following convention for probability distributions:
if the arguments of p( ) are different, they are different
functions, thus: p(x) ≠ p(y) even if y = x.
12. One can prove the existence of a metric directly but in order
to perform the metrization the space has to be decomposed first.
But decomposing a space without having a metric calls for a neat
trick, accomplished (as far as we know) only by the method used
by the SOM.
12a. In this example we use a hemisphere; in general, it would be
a spherical cap.
A Topological Foundation for Self-Organization
R. I. ŚCIBOR-MARCHOCKI
_Northrop Nortronics_
_Systems Support Department_
_Anaheim, California_
It is shown that by the use of Information
Theory, any metrizable topology may be metrized
as an orthogonal Euclidean space (with a random
Gaussian probability distribution) times
a denumerable random cartesian product of
irreducible (wrt direct product) denumerable
groups. The necessary algorithm to accomplish
this metrization from a statistical basis is
presented. If such a basis is unavailable,
a certain nilpotent projection operator has
to be used instead, as is shown in detail in
the companion paper. This operator possesses
self-organizing features.
INTRODUCTION
In the companion article[8] we will define a self-organizing system
as one which, after observing the input and output of an unknown
phenomenon (transfer relation), organizes itself into a simulation of
the unknown phenomenon.
[8] Kleyn, P. A., “Conceptual Design of Self-Organizing Machines,”
Anaheim, California:Northrop Nortronics, NSS Report 2832, Nov. 14, 1963.
Within the mathematical model, the aforementioned phenomenon may be
represented as a topological space thus omitting for the moment the
(arbitrary) designation of input and output which, as will be shown,
bears on the question of uniqueness. Hence, for the purpose of this
paper, which emphasizes the mathematical foundation, an intelligent
device is taken as one which carries out the task of studying a space
and describing it.
In keeping with the policy that one should not ask someone (or
something) else to do a task that he could not do himself (at least in
principle), let us consider how we would approach such a problem.
In the first place, we have to select the space in which the problem is
to be set. The most general space that we feel capable of tackling is
a metrizable topology. On the other hand, anything less general would
be unnecessarily restrictive. Thus, we choose a metrizable topological
space.
As soon as we have made this choice, we regret it. In order to
improve the situation somewhat, we show that there is no (additional)
loss of generality in using an orthogonal Euclidean space times[9]
a denumerable random cartesian product of irreducible (wrt direct
product) denumerable groups.
This paper provides a survey of the problem and a method for solving
it which is conceptually clear but not very practical. The companion
paper[10] provides a practical method for solving this problem by means
of the successive use of a certain nilpotent projection operator.
[9] Random cartesian product.
[10] Kleyn, P. A., “Conceptual Design of Self-Organizing Machines,”
Anaheim, California:Northrop Nortronics, NSS Report 2832, Nov. 14, 1963.
METRIZATION
We start with a metrizable topological space. There are many equivalent
axiomatizations of a metrizable topology; _e.g._, see Kelley. Perhaps
the easiest way to visualize a metrizable topology is to consider that
one was given a metric space but that he lost his notes in which the
exact form of the metric was written down. Thus one knows that he can
do everything that he could in a metric space, if only he can figure
out how.
The “figuring out how” is by no means trivial. Here, it will be assumed
that a cumulative probability distribution has been obtained on the
space by one of the standard methods; bird in cage,[11] Munroe I,[12]
Munroe II,[13] ordering (see Halmos[14] or Kelley[15]). This cumulative
probability distribution is a function on X onto the interval [0,1] of
real numbers. The inverse of this function, which exists by the Radon
Nikodym theorem, provides a mapping from the real interval onto the
non-trivial portion of X. This mapping induces all of the pleasant
properties of the real numbers on the space X: topological, metric, and
ordering.
Actually, it turns out that, especially if the dimensionality of
the space is greater than one, the foregoing procedure not only
provides one metrization, but many. Indeed, this lack of uniqueness
is what makes the procedure exceedingly difficult. Only by imposing
some additional conditions that result in the existence of a unique
solution, does the problem become tractable.
We choose to impose the additional condition that the resulting metric
space be a Euclidean geometry with a rectangular coordinate system.
[11] Harman, W. W., “Principles of the Statistical Theory of
Communication,” New York, New York:McGraw-Hill, 1963.
[12] Munroe, M. E., “Introduction to Measure and Integration,”
Cambridge, Mass.:Addison-Wesley, 1953.
[13] Munroe, M. E., “Introduction to Measure and Integration,”
Cambridge, Mass.:Addison-Wesley, 1953.
[14] Halmos, P. R., “Measure Theory,” Princeton, New Jersey:D. Van
Nostrand Co., Inc., 1950.
[15] Kelley, J. L., “General Topology,” Princeton, New Jersey:D. Van
Nostrand Co., Inc., 1955.
Even this always does not yield uniqueness, but we will show the
additional restriction that will guarantee uniqueness after the
necessary language is developed. Since all metrizations of a given
metrizable topology are isomorphic, in the quotient class the
orthogonal Euclidean geometry serves the purpose of being a convenient
representative of the unique element resulting from a given metrizable
topology.
Furthermore, the same comment applies to the use of a Gaussian
distribution as the probability distribution on this orthogonal
Euclidean geometry. Namely, the random Gaussian distribution on an
orthogonal Euclidean geometry is a convenient representative member of
the equivalence class which maps into one element (stochastic space) of
the quotient class.
Information Theory
Now, we will show that Information Theory provides the language
necessary to describe the metrization procedure in detail.
It is possible to introduce Information Theory axiomatically by a
suitable generalization of the axioms[16] in Feinstein.[17] But
to simplify the discussion here, we will use the less elegant but
equivalent method of defining certain definite integrals. The
probability density distribution p is defined from the cumulative
probability distribution P by
P(X′) = ∫X′_{measurable ⊂ X} p(x)dx. (1)
Then the information rate H is defined as
H(X) = -∫ₓp(x) ln κ p(x)dx (2)
where kappa has (carries) the units of X. Finally,
the channel rate R is defined as
R(⨀Xᵢ) = ΣH(Xᵢ) - H(X), (3)
I I
where X is the denumerable[18] cartesian product space
X = ⨂Xᵢ. (4)
I
[16] Feinstein uses his axioms only in finite space X; _i.e._, card(X)
< K₀.
[17] Feinstein, A., “Foundations of Information Theory,” New York, New
York: McGraw-Hill, 1958.
[18] If I is infinite, certain precautions have to be exercised.
Next, we define the angle Θ
|Θ(⨀Xᵢ)| = sin⁻¹_e_^{-R(⨀Xᵢ)} (5)
I
and the norm
|X| = κ(2π_e_)⁻¹ᐟ² _e_^{(HX)}. (6)
Now, if[19] a statistically independent basis; _i.e._, one for which
κ
R(⨀Xᵢ) ≡ constant, (7)
I
can be provided in terms of one-dimensional components; _i.e._, none
of them can be decomposed further, then it is just the usual problem
of diagonalization of a symmetric matrix by means of a congruence
transformation to provide an orthogonal coordinate system. Furthermore,
for uniqueness, we arrange the spectrum in decreasing order. Then,
by means of the Radon Nikodym theorem applied to each of these
one-dimensional axes, the probability distribution may be made; _e.g._,
Gaussian, if desired. Thus, we obtain the promised orthogonal Euclidean
space.
[19] This “if” is the catch that makes all methods of metrization of a
space of dimensionality higher than one impractical, except the method
of successive projections upon unit spheres centered at the center of
gravity. The method of using that nilpotent projection operator is
described in the companion paper(see footnote page 65).
Channel
At this time we can state the remaining additional condition required
that a decomposition be unique. The index space I has to be partitioned
into exactly two parts, say I′ and I″; _i.e._,
I′ ∪ I″ = I (8)
I′ ∩ I″ = φ,
such that
dim(X′) = dim(X″), (9)
where
X′ = ⨂Xᵢ (10)
I′
X″ = ⨂Xᵢ.
I″
(If dim (X) is odd, then we have to cheat a little by putting in an
extra random dummy dimension.) And then the decomposition of the space
X = ⨂Xᵢ (11) I
has to be carried out so that this partitioning is preserved.
Since this partitioning is arbitrary (as far as the mathematics is
concerned), it is obvious that a space which is not partitioned will
have many (equivalent) decompositions. On the other hand, if the
partitioning is into more than two parts, then the existence of a
decomposition is not guaranteed.
A slight penalty has to be paid for the use of this partitioning,
namely: instead of eventually obtaining a random cartesian product of
one-dimensional spaces, we obtain an extended channel (with random
input) of single-dimensional channels. It is obvious that if we were
to drop the partitioning temporarily, each such single-dimensional
channel would be further decomposed into two random components. This
decomposition is not unique. But one of these equivalent decompositions
is particularly convenient; namely, that decomposition where we take
the component out of the original X′ and that which is random to it,
say V. This V (as well as the cartesian product of all such V’s,
which of necessity are random) is called the linearly additive noise.
The name “linearly additive” is justified because it is just the
statistical concept isomorphic to the linear addition of vectors in
orthogonal Euclidean geometry. (The proof of this last statement is not
completed as yet.)
Denumerable Space
The procedure for this decomposition was worded to de-emphasize the
possible presence of a denumerable (component of the) space. Such a
component may be given outright; otherwise, it results if the space was
not simply connected. Any denumerable space is zero dimensional, as may
be verified easily from the full information theoretic definition of
dimensionality.
The obvious way of disposing of a denumerable space is to use the
conventional mapping that converts a Stieltjes to a Lebesque integral,
using fixed length segments. (It can be shown that H is invariant
under such a mapping.) Unfortunately, while this mapping followed by
a repetition of the preceding procedure will always solve a given
problem (no new[20] denumerable component _need_ be generated on the
second pass), little insight is provided into the structure of the
resulting space. On the other hand, because channels under cascading
constitute a group, any such denumerable space is a representation of a
denumerable group.
[20] Only non-cyclic irreducible (wrt direct product) denumerable group
components of the old denumerable space will remain.
SUMMARY
In summary, the original metrizable topological space was decomposed
into an orthogonal Euclidean space times[21] a denumerable random
cartesian product of irreducible (wrt direct product) denumerable
groups. Thus, since any individual component of a random cartesian
product may be studied independently of the others, all that one needs
to study is: (1) a Gaussian distribution on a single real axis and (2)
the irreducible denumerable groups.
[21] Random cartesian product.
Finally, it should be emphasized that there are only these two ways
of decomposing a metrizable topology; (1) if a (statistical) basis
is given, use the diagonalization of a symmetric matrix algorithm
described earlier (and given in detail in the three channels in cascade
problem), and (2) otherwise use a suitable network of the NPO’s with
n₀=1. Of course, any hybrid of these two methods may be employed as
well.
On Functional Neuron Modeling
C. E. HENDRIX
_Space-General Corporation_
_El Monte, California_
There are two very compelling reasons why mathematical and physical
models of the neuron should be built. Model building, while widely
used in the physical sciences, has been largely neglected in biology.
However, there can be little doubt that building neuron models
will increase our understanding of the function of real neurons,
if experience in the physical sciences is any guide. Secondly,
neuron models are extremely interesting in their own right as new
technological devices. Hence, the interest in, and the reason for
symposia on self-organizing systems.
We should turn our attention to the properties of real neurons, and
see which of them are the most important ones for us to imitate.
Obviously, we cannot hope to imitate _all_ the properties of a living
neuron, since that would require a complete simulation of a living,
metabolizing cell, and a highly specialized one at that; but we
can select those functional properties which we feel are the most
important, and then try to simulate those.
The most dramatic aspect of neuron function is, of course, the axon
discharge. It is this which gives the neuron its “all-or-nothing”
character, and it is this which provides it with a means for
propagating its output pulses over a distance. Hodgkin and Huxley (1)
have developed a very complete description of this action. Their model
is certainly without peer in describing the nature of the real neuron.
On the technological side, Cranes’ “neuristors” (2) represent a class
of devices which imitate the axonal discharge in a gross sort of way,
without all the subtle nuances of the Hodgkin-Huxley model. Crane has
shown that neuristors can be combined to yield the various Boolean
functions needed in a computer.
However, interesting as such models of the axon are, there is some
question as to their importance in the development of self-organizing
systems. The pulse generation, “all-or-nothing” part of the axon
behavior could just as well be simulated by a “one-shot” trigger
circuit. The transmission characteristic of the axon is, after all,
only Nature’s way of sending a signal from here to there. It is
an admirable solution to the problem, when one considers that it
evolved, and still works, in a bath of salt water. There seems little
point, however, in a hardware designer limiting himself in this way,
especially if he has an adequate supply of insulated copper wire.
If the transmission characteristic of the axon is deleted, the
properties of the neuron which seem to be the most important in the
synthesis of self-organizing systems are:
a. The neuron responds to a stimulus with an electrical
pulse of standard size and shape. If the stimulus
continues, the pulses occur at regular intervals
with the rate of occurrence dependent on the
intensity of stimulation.
b. There is a threshold of stimulation. If the
intensity of the stimulus is below this threshold,
the neuron does not fire.
c. The neuron is capable of temporal and spatial
integration. Many subthreshold stimuli arriving at
the neuron from different sources, or at slightly
different times, can add up to a sufficient level to
fire the neuron.
d. Some inputs are excitatory, some are inhibitory.
e. There is a refractory period. Once fired, there is
a subsequent period during which the neuron cannot
be fired again, no matter how large the stimulus.
This places an upper limit on the pulse rate of any
particular neuron.
f. The neuron can learn. This property is conjectural
in living neurons, since it appears that at
the present time learning has not been clearly
demonstrated in isolated living neurons.
However, the learning property is basic to all
self-organizing models.
Neuron models with the above characteristics have been built, although
none seem to have incorporated _all_ of them in a single model. Harman
(3) at Bell Labs has built neuron models which have the characteristics
(a) through (e), with which he has built extremely interesting devices
which simulate portions of the peripheral neuron system.
Various attempts at learning elements have been made, perhaps best
exemplified by those of Widrow (4). These devices are capable of
“learning,” but are static, and lack all the temporal characteristics
listed in (a) through (e). Such devices can be used to deal with
temporal patterns only by a mapping technique, in which a temporal
pattern is converted to a spatial one.
Having listed which seem to be the important properties of a neuron, it
is possible to synthesize a simple model which has all of them.
A number of input stimuli are fed to the neuron through a resistive
summing network which establishes the threshold and accomplishes
spatial integration. The voltage at the summing junction triggers a
“one-shot” circuit, which, by its very nature, accomplishes pulse
generation and exhibits temporal integration and a refractory period.
The polarity of an individual input determines whether it shall be
excitatory or inhibitory. This much of the circuitry is very similar to
Harmon’s model.
Learning is postulated to take place in the following way: when the
neuron fires, an outside influence (the environment, or a “trainer”)
determines whether or not the result of firing was desirable or not.
If it was desirable, the threshold of the neuron is lowered, making
it easier to fire the next time. If the result was not desirable, the
threshold is raised, making it more difficult for the neuron to fire
the next time.
In a self-organizing system, many model neurons would be
interconnected. A “punish-reward” (P-R) signal would be connected to
all neurons in common. However, means would be provided for only those
which have recently fired to be susceptible to the effects of the P-R
signal. Therefore, only those which had taken part in a recent response
are modified. This idea is due to Stewart (5), who applies it to his
electrochemical devices instead of to an electronic device.
The mechanization of the circuitry is rather straight-forward. A
portion of the output of the pulse generator is routed through a
“pulse-stretcher” or short-term memory which temporarily records the
fact that the neuron has recently fired. The pulse-stretcher output
controls a gate, which either accepts or rejects the P-R signal. The
P-R signal can take on only three values, a positive level, zero, or
a negative level, depending on whether the signal is “punish,” “no
action,” or “reward.” Finally, the gate output controls a variable
resistor, which is part of the resistive summing network. Figure 1 is a
block diagram of the complete model.
Note that this device differs from the usual “Perceptron” configuration
in that the threshold resistor is the only variable element, instead
of having each input resistor a variable weighting element. This
simplification could lead to a situation where, to prepare a specified
task, more single-variable neurons would be required than would
multivariable ones. This possible disadvantage is partially, at least,
offset by the very simple control algorithm which is contained in the
design of the model, and is not the matter of great concern which it
seems to be for most multivariable models.
[Illustration: Figure 1—Block diagram of neuron model]
Hand simulations of the action of this type of model suggest that a
certain amount of randomness would be desirable. It appears that a
self-organizing system built of these elements, and of sufficient
complexity to be interesting, would have a fair number of recirculating
loops, so that spontaneous activity would be maintained in the absence
of input stimulus. If this is the case, then randomness could easily
be introduced by adding a small amount of noise from a random noise
generator to the signal on the P-R bus. Thus, any neurons which
spontaneously fire would be continually having their thresholds
modified.
The mechanization of the model is not particularly complex, and can
be estimated as follows: The one-shot pulse generator would require
two transistors, the pulse stretcher one more. The bi-directional gate
would require a transistor and at least two diodes.
Several candidates for the electrically-controllable variable
resistor are available (6). Particularly good candidates appear to
be the “Memistor” or plating cell developed by Widrow (7), the solid
state version of it by Vendelin (8), and the “solion” (9). All are
electrochemical devices in which the resistance between two terminals
is controlled by the net charge flow through a third terminal. All are
adaptable to this particular circuit.
Of the three, however, the solion appears at first glance to have the
most promise in that its resistance is of the order of a few thousand
ohms (rather than the few ohms of the plating cells) which is more
compatible with ordinary solid-state circuitry. Solions have the
disadvantage that they can stand only very low voltages (less than 1
volt) and in their present form require extra bias potentials. If these
difficulties can be overcome, they offer considerable promise.
In summary, it appears that a rather simple neuron model can be built
which can mimic most of the important functions of real neurons. A
system built of these could be punished or rewarded by an observer,
so that it could be trained to give specified responses to specified
stimuli. In some cases, the observer could be simply the environment,
so that the system would learn directly from experience, and would be
therefore a self-organizing system.
REFERENCES
1. Hodgkin, A. L., and Huxley, A. L.,
“A Quantitative Description of Membrane Current and its
Application to Conduction and Excitation in Nerve,”
_J. Physiol._ =117=:500-544 (August 1952)
2. Crane, H. D.,
“Neuristor—A Novel Device and System Concept,”
_Proc. IRE_ =50=:2048-2060 (Oct. 1962)
3. Harmon, L. D., Levinson, J., and Van Bergeijk, W. A.,
“Analog Models of Neural Mechanism,”
_IRE Trans. on Information Theory_ =IT-8=:107-112
(Feb. 1962)
4. Widrow, B., and Hoff, M. E.,
“Adaptive Switching Circuits,”
Stanford Electronics Lab Tech Report 1553-1, June 1960
5. Stewart, R. M.,
“Electrochemical Wave Interactions and Extensive Field Effects
in Excitable Cellular Structures,”
First Pasadena Invitational Symposium on Self-Organizing Systems,
Calif. Institute of Technology, Pasadena, Calif., 14 Nov. 1963
6. Nagy, G.,
“A Survey of Analog Memory Devices,”
_IEEE Trans. on Electronic Cmptrs._ EC-12:388-393 (Aug. 1963)
7. Widrow, B.,
“An Adaptive Adaline Neuron Using Chemical Memistors,”
Stanford Electronics Lab Tech Report 1553-2, Oct. 1960
8. Vendelin, G. D.,
“A Solid State Adaptive Component,”
Stanford Electronics Lab Tech Report 1853-1, Jan. 1963
9. “Solion Principles of Electrochemistry and Low-Power
Electrochemical Devices,”
Dept. of Comm., Office of Tech. Serv. =PB= 131931
(U. S. Naval Ord. Lab., Silver Spring, Md., Aug. 1958)
Selection of Parameters for Neural Net Simulations[22]
R. K. OVERTON
_Autonetics Research Center_
_Anaheim, California_
Research of high quality has been presented at this Symposium. Of
particular interest to me were the reports of the Aeronutronic group
and the Librascope group. The Aeronutronic group was commendably
systematic in its investigations of different arrangements of linear
threshold elements, and the Librascope data, presenting the effects of
attaching different values to the parameters of simulated neurons, are
both systematic and interesting.
Unfortunately, however, interest in such research can obscure a more
fundamental question which seems to merit study. That question concerns
the parameters, or attributes, which describe the simulated neuron.
Specifically, which parameters or attributes should be selected for
simulation? (For example, should a period of supernormal sensitivity be
simulated following an absolutely refractory period?)
Some selection obviously has to be made. Librascope, which is
trying to simulate neurons more or less faithfully, plans to build
a net of ten simulated neurons. In contrast, General Dynamics/Fort
Worth, with roughly the same degree of effort, is working with 3900
unfaithfully-simulated neurons. This comparison is not a criticism
of either group; the Librascope team has simply selected many more
parameters for simulation than has the General Dynamics group. Each
can make the selections it prefers, because the parameters of real
neurons which are necessary and sufficient for learning have not been
exhaustively identified.
From the point of view of one whose interests include real neurons,
this lack of identification is unfortunate. I once wrote a book which
included some guesses about the essential attributes of neurons. Since
that time, many neuron simulation programs have been written. But these
programs, although interesting and worthwhile in their own right, have
done little to answer the question of the necessary parameters. That
is, they do not make much better guesses possible. And yet better
guesses would also make for more “intelligent” machines.
[22] This paper, submitted after the Symposium, represents a more
detailed presentation of some of the issues raised in the discussion
sessions at the Symposium and hence, constitutes a worthwhile addition
to the Proceedings.
INDEX OF INVITED PARTICIPANTS
MICHAEL ARBIB Massachusetts Institute of Technology
ROBERT H. ASENDORF Hughes Research Laboratories/ Malibu
J. A. DALY Astropower/Newport Beach
GEORGE DeFLORIO System Development Corp./Santa Monica
DEREK H. FENDER California Institute of Technology
LEONARD FRIEDMAN Space Technology Labs./Redondo Beach
JAMES EMMETT GARVEY ONR/Pasadena
THOMAS L. GRETTENBERG California Institute of Technology
HAROLD HAMILTON Librascope/Glendale
JOSEPH HAWKINS Aeronutronic/Newport Beach
CHARLES HENDRIX Space-General Corp./El Monte
R. D. JOSEPH Astropower/Newport Beach
PETER A. KLEYN Nortronics/Anaheim
JOHN KUHN Space-General Corp./El Monte
FRANK LEHAN Space-General Corp./El Monte
EDWIN LEWIS Librascope/Glendale
PETER C. LOCKEMANN California Institute of Technology
GILBERT D. McCANN California Institute of Technology
C. J. MUNCIE Aeronutronic/Newport Beach
C. OVERMIER Nortronics/Anaheim
RICHARD K. OVERTON Autonetics/Anaheim
DIANE RAMSEY Astropower/Newport Beach
RICHARD REISS Librascope/Glendale
R. I. ŚCIBOR-MARCHOCKI Nortronics/Anaheim
JAMES J. SPILKER Philco/Palo Alto
ROBERT M. STEWART Space-General Corp./El Monte
HENNIG STIEVE California Institute of Technology
RICHARD TEW Space-General Corp./El Monte
JOHN THORSEN University of California/Los Angeles
RICHARD VINETZ Librascope/Glendale
CHRISTOPH von CAMPENHAUSEN California Institute of Technology
DAVID VOWLES California Institute of Technology
HORST WOLF Astropower/Newport Beach
U.S. GOVERNMENT PRINTING OFFICE: 1966 O—205-502
*** End of this LibraryBlog Digital Book "Self-Organizing Systems, 1963" ***
Copyright 2023 LibraryBlog. All rights reserved.