人工智能纳米光子学:光学神经网络与纳米光子学

Artificial Intelligence Nanophotonics: Optical Neural Networks and Nanophotonics

人工智能技术,特别是人工神经网络的创新引领了许多领域的应用革命,如网络搜索、计算机识别和语言、图像的识别技术。近年来纳米光子学的发展为传统的人工神经网络技术,特别是光学神经网络的发展带来了全新的物理视角以及截然不同的实现方法。一方面,纳米光子学是一门研究光与材料在纳米尺度相互作用的科学,可以带来全新的技术,如超分辨光学加工技术和超分辨光学成像技术,进而推动微纳尺度上多种功能的光学神经网络的实现。另一方面,纳米光子学中光子传播的多频段、高速度、低功耗的特点,促使了光学神经网络向着小体积、高密度、低功耗的方向发展。人工神经网络自身的发展也促使神经网络算法(如逆向设计、深度学习)在纳米光子学器件的设计中发挥前所未有的作用,以满足纳米光子学器件对自身功能、体积、集成度、计算功能的日益增长的要求。以神经网络的发展为起点,阐述人工神经网络特别是光学神经网络的发展趋势,以及人工神经网络与纳米光子学相互促进的发展历程。

Innovations in artificial intelligence,particularly artificial neural networks,have revolutionized applications in many areas,such as big-data search,computer recognition,and language and image recognition.The development of nanophotonics in the past decades has brought physical perspectives and different approaches to the implementation and the development of traditional artificial neural network technologies,especially optical neural networks.On the one hand,nanophotonics is a field studying the interaction of light and matter at the nanoscale,which can lead to new techniques,such as super-resolution optical lithography and super-resolution optical imaging technology,therefore in turn promoting the implementation of optical neural networks with multiple functions at the micro/nano scale.On the other hand,due to the characteristics of multi-bands,high speed,and low power consumption of light propagation,nanophotonics is accelerating the development of optical neural networks with compact size,high density,and low power consumption.Meanwhile,the development of artificial neural networks has also promoted neural network algorithms(such as reverse design and deep learning)as a new toolbox for the design of novel nanophotonics devices to meet the growing requirements of the function,volume,integration,and computing function of nano-photonic devices.In this paper,starting with the development of neural networks,we review the development of artificial neural networks,especially the development of optical neural networks.The reciprocal development between artificial neural networks and nanophotonics is reviewed.