Exploitation des propriétés de saturation synaptique pour obtenir un neurone à fréquence spécifique

Abstract

Energy consumption remains the main limiting factors in many promising IoT applications. In particular, micro-controllers consume far too much power. In order to overcome this problem, new circuit designs have been proposed and the use of spiking neurons and analog computing has emerged as it allows a very significant consumption reduction. However, working in the analog domain brings difficulty to handle the sequential processing of incoming signals as is needed in many use cases.In this paper, we propose to use a bio-inspired phenomenon called Interacting Synapses to produce a time filter. We propose a model of synapses that makes the neuron fire for a specific range of delays between two incoming spikes, but not react when this Inter-Spike Timing is not in that range. We study the parameters of the model to understand how to adapt the Inter-Spike Timing. The originality of the paper is to propose a new way, in the analog domain, to deal with temporal sequences.

Tasks

Reproductions