Approximation of Spike-trains by Digital Spiking Neuron

A digital spiking neuron (DSN) consists of shift registers and can generate spike-trains with various patterns of inter-spike intervals. In this paper we present a learning algorithm for the DSN in order to approximate given spike-trains. We study a case where a student DSN accepts a spike-train from a teacher DSN. It is shown that the student can reproduce a spike-train of the teacher based on the learning algorithm. We also study a case where a chaotic analog spiking neuron is used as a teacher. It is shown that the DSN can approximate a sampled chaotic spike-train with a small error.