Memory trace in spiking neural networks

Abstract

Spiking neural networks have a limited memory capacity, such that a stimulus arriving at time t would vanish over a timescale of 200-300 milliseconds [1]. Therefore, only neural computations that require history dependencies within this short range can be accomplished. In this paper, the limited memory capacity of a spiking neural network is extended by coupling it to an delayed-dynamical system. This presents the possibility of information exchange between spiking neurons and continuous delayed systems. © 2013 Springer-Verlag Berlin Heidelberg.