Wave-Based Spiking Neural Network with Nano-Structured Electronics

Abstract

The present paper is on neuromorphic computing with temporal spike signaling as elastic waves and building blocks using nano-structured electronic systems. Wave-based neuromorphic computing framework for energy-efficient brain-inspired computing [1] can natively calculate weighted-sums with delay spreads as wave superposition and propagation for temporal spiking neural networks (SNN) processing. Single particle quantum effects of electron wave packets, traveling at the Fermi velocity in nano-structured systems, are exploited to constitute various building blocks. We will first describe passive elements whose delay can be widely controlled electrically as well as geometrically. Then, splitter combiner building blocks are configured with coupled quantum systems and an alignment gate. Finally, we explain self-resetting integrate-and-fire building blocks with tunneling in coupled nano systems. Independent changes of the static potential at subband edges and the Fermi potential lead to rich dynamics of Coulomb and quantum interplay to accomplish critical SNN building block functions.