A mathematical design method for silicon neuron circuit

Abstract

    Silicon neuron is electrical circuit that mimics the electrophysiological functions of a neuronal cell. One of the goals of researches on silicon neuron is to construct some functional silicon neural networks by connecting silicon neurons each other via silicon synapses that mimics the electrophysiological function of synapse. It is expected to realize an intelligent artificial system that operates based on novel principles similar to brain and nerve systems. To construct such systems, a silicon neuron circuit has to be simple and consumes low power. Some researchers are designing silicon neuron by analog electrical circuit that solves differential equations of ionic-conductance models of a neuronal cell. Such circuits are proved to reproduce various behaviors of the neuron model quite precisely in real time. However, because the ionic-conductance models describe the behavior of lipid and protein, they are not quite compatible to the characteristics of electrical circuit, which restricts the simplification of circuitry and reduction of power consumption. To solve this problem, we proposed a new designing method that is based on mathematical knowledge on ionic-conductance models and designed and implemented a new type of silicon neuron circuit.