Modern machine-learning applications require huge artificial networks demanding computational power and memory.Light-based platforms promise ultrafast and energy-efficient hardware,which may help realize next-generation data processing devices.However,current photonic networks are limited by the number of input-output nodes that can be processed in a single shot.This restricted network capacity prevents their application to relevant large-scale problems such as natural language processing.Here,we realize a photonic processor for supervised learning with a capacity exceeding 1.5×1010 optical nodes,more than one order of magnitude larger than any previous implementation,which enables photonic large-scale text encoding and classification.By exploiting the full three-dimensional structure of the optical field propagating in free space,we overcome the interpolation threshold and reach the over-parameterized region of machine learning,a condition that allows high-performance sentiment analysis with a minimal fraction of training points.Our results provide a novel sol-ution to scale up light-driven computing and open the route to photonic natural language processing.

CARLO M.VALENSISE;IVANA GRECCO;DAVIDE PIERANGELI;CLAUDIO CONTI

Enrico Fermi Research Center(CREF),00184 Rome,Italy;Physics Department,Sapienza University of Rome,00185 Rome,Italy;Institute for Complex Systems,National Research Council(ISC-CNR),00185 Rome,Italy

光子学研究（英文）

[1]CARLO M.VALENSISE;IVANA GRECCO;DAVIDE PIERANGELI;CLAUDIO CONTI-.Large-scale photonic natural language processing)[J].光子学研究（英文）,2022(12):2846-2853

Large,scale,photonic,natural,language,processing,Modern,machine,learning,applications,require,huge,artificial,networks,demanding,computational,power,memory,Light,platforms,promise,ultrafast,energy,efficient,hardware,which,may,help,realize,next,generation,data,devices,However,current,limited,by,number,input,output,nodes,that,can,processed,single,shot,This,restricted,capacity,prevents,their,relevant,problems,such,Here,processor,supervised,exceeding,optical,more,than,one,order,magnitude,larger,any,previous,implementation,enables,text,encoding,classification,By,exploiting,full,three,dimensional,structure,field,propagating,free,space,overcome,interpolation,threshold,reach,parameterized,region,condition,allows,high,performance,sentiment,analysis,minimal,fraction,training,points,Our,results,provide,novel,sol,ution,light,driven,computing,open,route

0.710426