GRODEC, THE GRODEC ACTION POTENTIAL, AND GRODEC-BASED NEURAL NETWORKS
Date
1995-12-01
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Grodecs are abstract growth decay entities that are related to
oscillators. The switching behaviour of grodecs has been analysed
in detail and it is shown that grodecs can be used as the basis of
both an unconventional type of neural net called a grocec stack machine,
as well as conventional neural nets.
A single grodec has the most of the switching properties of a short section
of biological membrane, and can give rise to action potentials when
properly set. A pair of grodecs can generate a much sharper action potential
than can a single grodec, and this mechanism is very like that of a neuron
with dual Na and K ion switching, so that a neuron membrane can be
regarded conceptually as a Na grodec coupled to a K grodec. Grodecs can be
coupled in positive (excitatory) or negative (inhibitory) manner to form
logic circuits. Examples of such circuits are given, and show that grodecs
can be the basis of conventional neural nets and sequential machines in
general. Grodecs can also be coupled in an unconventional manner to form
grodec stack machines. In such machines the action potentials sum in periods
of growth and decay, so that a central machine pressure parameter exhibits
time function behaviour similar to that of a fractal growth-decay time
function, such as an En:m function. Although the manner of the summation is
much simpler than in grodec stack machines, action potential summation does
occur in nature, and causes the large voltages generated by electric fish.
Description
Keywords
Computer Science