Bifurcation analysis of a reconfigurable hybrid spiking neuron and its novel online learning algorithm

  • Authors:
  • Sho Hashimoto;Hiroyuki Torikai

  • Affiliations:
  • Graduate School of Engineering Science, Osaka University, Japan;Graduate School of Engineering Science, Osaka University, Japan

  • Venue:
  • IJCNN'09 Proceedings of the 2009 international joint conference on Neural Networks
  • Year:
  • 2009

Quantified Score

Hi-index 0.00

Visualization

Abstract

A hybrid spiking neuron is a wired system of shift registers and behaves like a neuron model. The neuron exhibits various bifurcation phenomena and response characteristics for a stimulation spike-train input. In this paper we formulate some typical bifurcation mechanisms and clarify these mechanisms by using discrete/continuous hybrid maps. Based on the analysis results, we can clarify mechanisms of various responses of the neuron. In addition, we propose a novel online learning algorithm of the neuron and show that the neuron can reconstruct or approximate response characteristics of another neuron with unknown parameter values.