Approximation of unknown neuron dynamics by a discrete-state spiking neuron model: an approach toward neural prosthesis

摘要

In this paper, we introduce a hybrid spiking neuron which is a discrete-state neuron model. The neuron is suitable for a hardware implementation, since the neuron is a wired system of shift registers and the control parameters are wiring patterns among them. We analyze various responses of the neuron to a stimulation input. In addition, we present a learning algorithm for the neuron. The learning algorithm enables the neuron to reconstruct or approximate the response characteristics of anther neuron with unknown parameter values. We also propose a learning method for a connected system of two neurons and analyze its basic characteristics.