Intelligent neuromorphic computing based on nanophotonics and metamaterials

Abstract

With the rapid development of artificial intelligence and computing chips approaching the bottleneck of power consumption and computing power, the research on intelligent computing hardware with high speed and high energy efficiency is an important trend. Recently, neuromorphic computing represented by photonic circuit neural networks and all-optical diffraction neural networks has attracted widespread attention due to their ultra-fast and ultra-efficient computing architectures. In this perspective, we first review some representative works and introduce them through two main lines of planar photonic circuit neural networks and three-dimensional diffraction neural networks to compare their characteristics and performance. We further discuss programmable designs for neuromorphic computing hardware, which bring it closer to general-purpose computing devices. Besides intelligent neural networks in the optical band, we also review the development and application of the diffractive neural networks in the microwave band, showing their programmable capabilities. Finally, we present the future directions and development trends of intelligent neuromorphic computing and its potential applications in wireless communications, information processing, and sensing.