Professional Documents
Culture Documents
Pitts, W. Some Observations On The Simple Neuron Circuit
Pitts, W. Some Observations On The Simple Neuron Circuit
MATHEMATICAL BIOPHYSICS
VOLUME 4, 1942
SOME O B S E R V A T I O N S ON T H E S I M P L E N E U R O N CIRCUIT
WALTER PITTS
T H E UNIVERSITY OF CHICAGO
the value Y~o, sk could, of course, not be affected. If all zeroes between
s~ and sk disappeared for y~ =- yio, a state of complete activity would
result, in accordance with (1). The only other possibility is that there
be no zero between s~ and sk for y~ < Yio, but one is initiated, say at
sk,, for y~ = y~0. Now if there is to be no interval of complete activity
at all, then, precisely when the zero at sk, is formed, or Yk, = 0, and
not at all before, sk must cease to be a zero, so that Yk > 0. Now yk is
a linear function of Yk, for all values of the latter greater than or
equal to zero, since there is then no zero between them: consequently
if, as is clearly possible, we select a small Y'k satisfying 0 < Y'k < Yk,
y'~ will be the value of this function for some argument Y'k, which
exceeds zero, by continuity and monotonicity; and since Y'k > 0 , the
zero at sk is broken down for this value of yk also; and consequently
for a y'~, > 0, which is contrary to hypothesis. Hence (1) holds. (2)
follows immediately: if we have complete activity between s~ and sj
for a given range of values of y~, then all y~ for i < k < ] v a r y linear-
ly with y~; and it is only when one of these, say y~, diminishes (lin-
early) to zero that the complete activity is interrupted. Now if com-
plete activity were to be reestablished for a higher value of y~, then
Yk, being the same monotone function of y~, would be exceeded by
zero, which is not compatible with complete activity. By (1), the
zero formed at sk could not be removed in any other way.
It is clear, by the continuity of the value of y~+~ as a function of
y~, that the quantities ~o~, 0~, and the ranges are connected by the
important relations
L;(s~) : A~.~+j #o + (~§ ~j',
Uj(s~) -- A~.~+i O~ + ~,i (j).
F o r the pair 00, ~, in the case where sj -~ s~.. -- s~, as determined
thus, we shall use the permanent notation ~o~, #,~. In this case we
have
L~ (s~) -~ A 0o~ +/~.~,
LEMMA 2.
I f a circuit C has been in complete activity for a period qn time-
units in length, and ~ < 1, then the excitation of s~ at the time qn +
is given by
1 -- Aq
y~ (qn + ~) -- #, l----Z-~ + ~, Aq, (1)
W A L T E E PITTS 125
1 --A q
y~+j ( q n + j + ~) : <7i+~(~) + A.i.~+j ,ui 1 - - A + )l~ Aq. (2)
Y~-- 1 - - A '
and if this value for y~ puts C into complete activity, it will determine
a possible steady state of complete activity: i.e., this will be the case
126 NEURON CIRCUIT
if and only if
C'). .--<91~.
1--A
Clearly one, and only one of these conditions can be fulfilled by a giv-
en circuit. We shall consequently have the
COROLLARY.
When 0 < A < 1, the circuit C has a unique steady-state activity
WALTER t'ITTS 127
pattern. This will also be the case when A < - 1, exvept that here
no steady state of complete activity is possible.
Suppose now that A > 1. Here again, condition (B) is excluded,
and there is no steady state of complete activity; the other possibil-
ities may be expressed as
A"). -->ao~
1--A
C"). 1--A =
which are not incompatible, so that we may well have two possible
steady states, given by A ~o4 + / ~ and A 01 + ~4 respectively, both of
which contain at least one zero.
It may be desired to trace the course of activity in C from the
initial value ).4 to whatever steady state, if any, is finally reached.
This may be done without difficulty by considering which region con-
tains 2~, the initial value of y4. If this region contain a possible
steady-state value of y4, then, if it is F1 or Fa the steady state will
obviously be attained immediately, while if it is F2, ir will be ap-
proached asymptotically, in accordance with (1). If the region con-
raining 24 do not contain a steady state then, if it be F1 or F~ the value
of y~ will enter immediately and move through the adjacent regions
until the first one containing a steady state is entered, whereupon
that steady state is either attained immediately or asymptotically ac-
cordingly as this final region is F1 or F3, or is F~; while if the initial
region of Y4 be F~ then, if only one of F1, F~ contain a possible steady
state value, then that one will be entered, at a time determinable by
solving (1) for t with y4 -- ~ or 014 accordingly as t h e final region be
F.~ or F~; but if both/'1 and F~ contain possible steady states, y~ will
be described by (1) until that value y~ ~ 004 or O~ which occurs first
befalls, whereupon it goes into the steady state of F1 or Fz respec-
tively.
It will be worth while to conclude our discussion with a consider-
ation of the interesting and important case where A -- -+ 1. We no-
tice first that i f A - - -+ 1, then the conditions (A) and (C) of the
theorem become
A'"'). ~4<_-0
C""). ~4_---0,
at least one of which must hold, so that there always exists a steady
state for G with at least one zero in this case. The interesting case
128 NEURON CIRCUIT
LITERATURE
Householder, A. S. 1941a. "A Theory of Steady-State Activity in Nerve-Fiber
Networks: I. Definitions and Preliminary Lemmas." Bull. Math. B~ophysic4,
3, 63-69.
Householder, A. S. 1941b. "A Theory of Steady-State Activity in Nerve-Fiber
Networks: II. The Simple Circuit." Bull. Math. Biophysics, 3, 105-112.
Householder, A. S. 1941c. "A Theory of Steady-State Activity in Nerve-Fiber
Networks: III. The Simple Circuit in Complete Activity." Bull. Math. Bio-
physics, $, 137-140.
Householder, A. S. 1942. "A Theory of Steady-State Activity in Nerve-Fiber
Networks: IV. N Circuits with a Common Synapse." Bull. Math. Biophysics,
4, 7-14.