Integrate-and-fire Coding and Hodgkin-Huxley Circuits Employing Silicon Diodes

D. D. Coon
Microtronics Associates, 4516 Henry Street, Pittsburgh, PA 15213

A. G. U. Perera
Department of Physics and Astronomy, Georgia State University, Atlanta, GA 30303

ABSTRACT
Cooled silicon p⁺-n-n⁺ diodes are interesting components for implementation of massive neural networks and parallel asynchronous processors because of their remarkable neuronlike electrical properties, simple structure, noise immunity and extremely low power requirements. New ultralow power figures are reported. An integrate-and-fire model is used to explain the generation of neuronlike spiketrains and action potential overshoot which is reported for the first time. These neuronlike features are correlated with Hodgkin-Huxley neuron equivalent circuit diagrams. Mathematical modeling of device operation demonstrates that the integrate-and-fire mechanism gives rise to an iterative map which correlates successive time intervals in a spiketrain time series. Information coding is governed by the parametric dependence of the iterative map. This means of information coding is not commonly used in neural network modeling but it appears very natural from the point of view of nonlinear dynamics and real neurons which generate spiketrains. © 1989 Pergamon Press plc

Return to Dr. Perera's Homepage
Return to Physics & Astronomy Homepage