An optical neural chip for implementing complex-valued neural network

H Zhang,G Q Lo,Y Z Shi,Y Zhang,X D Jiang,F K Muhammad,M H Yung,A Laing,H Cai,R Santagati,X S Luo,M Gu,B Dong,D L Kwong,S Paesani,L C Kwek,A Q Liu,J Thompson

doi:10.1038/s41467-020-20719-7

Abstract

Complex-valued neural networks have many advantages over their real-valued counterparts. Conventional digital electronic computing platforms are incapable of executing truly complex-valued representations and operations. In contrast, optical computing platforms that encode information in both phase and magnitude can execute complex arithmetic by optical interference, offering significantly enhanced computational speed and energy efficiency. However, to date, most demonstrations of optical neural networks still only utilize conventional real-valued frameworks that are designed for digital computers, forfeiting many of the advantages of optical computing such as efficient complex-valued operations. In this article, we highlight an optical neural chip (ONC) that implements truly complex-valued neural networks. We benchmark the performance of our complex-valued ONC in four settings: simple Boolean tasks, species classification of an Iris dataset, classifying nonlinear datasets (Circle and Spiral), and handwriting recognition. Strong learning capabilities (i.e., high accuracy, fast convergence and the capability to construct nonlinear decision boundaries) are achieved by our complex-valued ONC compared to its real-valued counterpart.

Highlights

Complex-valued neural networks have many advantages over their real-valued counterparts
Conventional digital electronic computing platforms exhibit significant slowdown when executing algorithms using complex-valued operations because complex numbers have to be represented by two real numbers[7,13], which increases the number of MAC operations—the most frequently used and computationally expensive component of the neural network algorithms[14,15]
Each Mach–Zehnder interferometers (MZIs) consists of two beam splitter (BS)–phase shifter (PS) pairs

Summary

Introduction

Complex-valued neural networks have many advantages over their real-valued counterparts. Conventional digital electronic computing platforms exhibit significant slowdown when executing algorithms using complex-valued operations because complex numbers have to be represented by two real numbers[7,13], which increases the number of MAC operations—the most frequently used and computationally expensive component of the neural network algorithms[14,15]. To overcome these hurdles, it has been proposed that the computationally taxing task of implementing neural networks be outsourced to optical computing[16] which is capable of truly complex-valued arithmetic. Other topics related to optical neural networks include on-chip training[33], optical nonlinear activations[34,35,36] and various neural network architectures[37,38,39]

Methods

Results

Conclusion