Spiking Rate and Latency Encoding with Resonant Tunnelling Diode Neuron Circuits and Design Influences

Abstract

Neuromorphic computing, inspired by the functionality and efficiency of biological neural systems, holds promise for advancing artificial intelligence and computational paradigms. Resonant tunneling diodes (RTDs), thanks to their ability to generate neuronal dynamical responses, such as excitable spiking and refractoriness, have recently emerged as candidates for use as opto-electronic spiking neurons in novel neuromorphic computing hardware. This work explores the ability of RTD spiking neurons to deliver information encoding mechanisms analogous to those observed in biological neurons, specifically spike firing rate and spike latency encoding. We also report detailed experimental and numerical studies on the impact that the RTD mesa radius and circuit inductance and capacitance have on its spiking properties, providing useful information for future design of RTD-based neural networks optimised for ultrafast (>GHz rate) processing capabilities. Finally, we showcase the application of spike rate encoding in RTD neurons for the ultrafast reconstruction of a eight-level amplitude signal, effectively filtering out the noise.