Monolithically integrated asynchronous optical recurrent accelerator

Bo Wu; Haojun Zhou; Junwei Cheng; Wenkai Zhang; Shiji Zhang; Chaoran Huang; Dongmei Huang; Hailong Zhou; Jianji Dong; Xinliang Zhang

doi:10.1186/s43593-025-00084-y

Your Location：

Home >

Browse articles >

Monolithically integrated asynchronous optical recurrent accelerator

eLight 2025年5卷

Research Article | Updated：2025-12-01

- Monolithically integrated asynchronous optical recurrent accelerator
- Monolithically integrated asynchronous optical recurrent accelerator
- Affiliations：
- Author bio：
  
  hhailongzhou@hust.edu.cn
  jjjdong@hust.edu.cn
  kxlzhang@mail.hust.edu.cn
- Funds：
  
  National Key Research and Development Project of China;National Natural Science Foundation of China;Knowledge Innovation Program of Wuhan -Basic Research
- DOI：10.1186/s43593-025-00084-y
  中图分类号：
- 收稿：2024-10-27，
  
  修回：2025-01-17，
  
  录用：2025-03-04，
  
  网络出版：2025-05-06，
  
  纸质出版：2025-12
- Accepted：
Scan QR Code
Bo Wu, Haojun Zhou, Junwei Cheng, 等. Monolithically integrated asynchronous optical recurrent accelerator[J]. eLight, 2025,5.

Bo Wu, Haojun Zhou, Junwei Cheng, et al. Monolithically integrated asynchronous optical recurrent accelerator[J]. eLight, 2025, 5.
Bo Wu, Haojun Zhou, Junwei Cheng, 等. Monolithically integrated asynchronous optical recurrent accelerator[J]. eLight, 2025,5. DOI： 10.1186/s43593-025-00084-y.

Bo Wu, Haojun Zhou, Junwei Cheng, et al. Monolithically integrated asynchronous optical recurrent accelerator[J]. eLight, 2025, 5. DOI： 10.1186/s43593-025-00084-y.

摘要

Abstract

Computing with light is widely recognized as a promising paradigm for overcoming the energy and latency limitations of electronic computing. However

the energy consumption and latency in current optical computing hardware predominantly arise in the electrical domain rather than the optical domain

primarily due to frequent signal conversions between optical (analog) and electrical (digital) formats. Furthermore

as the operating frequency of optical computing surpasses the GHz range

the synchronization of parallel electrical signals and the management of optical delays become increasingly critical. These challenges exacerbate energy consumption and latency

particularly in recurrent optical operations. To address these limitations

we propose a novel asynchronous computing paradigm for on-chip optical recurrent accelerators based on wavelength encoding

effectively mitigating synchronization challenges. By leveraging the intrinsic causality of wavelength relay

our approach eliminates the need for rigorous temporal alignment. To demonstrate the flexibility and efficacy of this asynchronous paradigm

we present two advanced recurrent models—an optical hidden Markov model and an optical recurrent neural network—monolithically integrated for the first time. These models incorporate hundreds of linear and nonlinear computing units densely packed into a compact footprint of just 10 mm

. Experimental evaluations on various benchmark tasks underscore the superior energy efficiency and low latency of the propose

d asynchronous optical accelerators. This innovation enables the efficient processing of large-scale parallel signals and positions optical processors as a pivotal technology for applications such as autonomous driving and intelligent robotics.

关键词

Keywords

references

H. Zhou et al. , Photonic matrix multiplication lights up photonic accelerator and beyond . Light Sci. Appl. 11 , 30 ( 2022 ). http://doi.org/10.1038/s41377-022-00717-8 http://doi.org/10.1038/s41377-022-00717-8

C. Li , X. Zhang , J. Li , T. Fang , X. Dong , The challenges of modern computing and new opportunities for optics . PhotoniX ( 2021 ). http://doi.org/10.1186/s43074-021-00042-0 http://doi.org/10.1186/s43074-021-00042-0

B. Wu , H. Zhou , J. Dong , X. Zhang , Programmable integrated photonic coherent matrix: principle, configuring, and applications . Appl. Phys. Rev. ( 2024 ). http://doi.org/10.1063/5.0184982 http://doi.org/10.1063/5.0184982

Y. Chen et al. , All-analog photoelectronic chip for high-speed vision tasks . Nature ( 2023 ). http://doi.org/10.1038/s41586-023-06558-8 http://doi.org/10.1038/s41586-023-06558-8

T. Wang et al. , Image sensing with multilayer nonlinear optical neural networks . Nat. Photonics ( 2023 ). http://doi.org/10.1038/s41566-023-01170-8 http://doi.org/10.1038/s41566-023-01170-8

S. Pai et al. , Experimental evaluation of digitally verifiable photonic computing for blockchain and cryptocurrency . Optica ( 2023 ). http://doi.org/10.1364/optica.476173 http://doi.org/10.1364/optica.476173

K. Vandoorne et al. , Experimental demonstration of reservoir computing on a silicon photonics chip . Nat. Commun. 5 , 3541 ( 2014 ). http://doi.org/10.1038/ncomms4541 http://doi.org/10.1038/ncomms4541

T. Zhou , W. Wu , J. Zhang , S. Yu , L. Fang , Ultrafast dynamic machine vision with spatiotemporal photonic computing . Sci. Adv. 9 , eadg4391 ( 2023 ). http://doi.org/10.1126/sciadv.adg4391 http://doi.org/10.1126/sciadv.adg4391

T. Honjo et al. , 100,000-spin coherent Ising machine . Sci. Adv. 7 , eabh0952 ( 2021 ). http://doi.org/10.1126/sciadv.abh0952 http://doi.org/10.1126/sciadv.abh0952

P.L. McMahon et al. , A fully programmable 100-spin coherent Ising machine with all-to-all connections . Science 354 , 614 - 617 ( 2016 ). http://doi.org/10.1126/science.aah5178 http://doi.org/10.1126/science.aah5178

T. Inagaki et al. , A coherent Ising machine for 2000-node optimization problems . Science 354 , 603 - 606 ( 2016 ). http://doi.org/10.1126/science.aah4243 http://doi.org/10.1126/science.aah4243

X. Liu , J. Cheng , H. Zhou , J. Dong , X. Zhang , Chip-scale all-optical complex-valued matrix inverter . APL Photonics ( 2024 ). http://doi.org/10.1063/5.0200149 http://doi.org/10.1063/5.0200149

D.C. Tzarouchis , M.J. Mencagli , B. Edwards , N. Engheta , Mathematical operations and equation solving with reconfigurable metadevices . Light Sci. Appl. ( 2022 ). http://doi.org/10.1038/s41377-022-00950-1 http://doi.org/10.1038/s41377-022-00950-1

Y.-W. Shen et al. , Deep photonic reservoir computing recurrent network . Optica ( 2023 ). http://doi.org/10.1364/optica.506635 http://doi.org/10.1364/optica.506635

Y. Shen et al. , Deep learning with coherent nanophotonic circuits . Nat. Photonics 11 , 441 - 446 ( 2017 ). http://doi.org/10.1038/nphoton.2017.93 http://doi.org/10.1038/nphoton.2017.93

H. Sun , K.-T. Wu , Enabling Technologies for High Spectral-Efficiency Coherent Optical Communication Networks 355 - 394 ( Wiley , London , 2016 ). http://doi.org/10.1002/9781119078289.ch10 http://doi.org/10.1002/9781119078289.ch10

L. Zhao , H. Dai , C. Yang , J. Lu , Y. Sun , Tradeoff between time and energy costs for controlling stochastic coupled neural networks . IEEE Trans. Autom. Control 69 , 1112 - 1118 ( 2024 ). http://doi.org/10.1109/tac.2023.3281471 http://doi.org/10.1109/tac.2023.3281471

S. Wang, M. Shi, D. Li and T. Du, A Survey of Time Synchronization Algorithms for Wireless Sensor Networks. 2019 Chinese Control Conference (CCC), Guangzhou, China, 2019, pp. 6392–6397. https://doi.org/10.23919/ChiCC.2019.8866385 10.23919/ChiCC.2019.8866385

F. Rivet et al. , The experimental demonstration of a SASP-based full software radio receiver . IEEE J. Solid State Circ. 45 , 979 - 988 ( 2010 ). http://doi.org/10.1109/jssc.2010.2041402 http://doi.org/10.1109/jssc.2010.2041402

B.S.G. Pillai et al. , End-to-end energy modeling and analysis of long-haul coherent transmission systems . J. Lightw. Technol. 32 , 3093 - 3111 ( 2014 ). http://doi.org/10.1109/jlt.2014.2331086 http://doi.org/10.1109/jlt.2014.2331086

M. Bahadori et al., Energy-performance optimized design of silicon photonic interconnection networks for high-performance computing. Design, Automation & Test in Europe Conference & Exhibition (DATE), 2017, Lausanne, Switzerland, 2017, pp. 326–331. https://doi.org/10.23919/DATE.2017.7927010 10.23919/DATE.2017.7927010

L.I. Gang , Z. Li-jun , L.I.N. Ling , H.E. Feng , Application of oversampling technique in upgrading ADC’s resolution in weak signal detection . Nanotechnol. Precis. Eng. 7 , 71 - 75 ( 2009 ).

M. Prabhu et al. , Accelerating recurrent Ising machines in photonic integrated circuits . Optica ( 2020 ). http://doi.org/10.1364/optica.386613 http://doi.org/10.1364/optica.386613

J. Cheng et al. , Multimodal deep learning using on-chip diffractive optics with in situ training capability . Nat. Commun. 15 , 6189 ( 2024 ). http://doi.org/10.1038/s41467-024-50677-3 http://doi.org/10.1038/s41467-024-50677-3

Z. Xu et al. , Large-scale photonic chiplet Taichi empowers 160-TOPS/W artificial general intelligence . Science 384 , 202 - 209 ( 2024 ). http://doi.org/10.1126/science.adl1203 http://doi.org/10.1126/science.adl1203

A.N. Tait et al. , Silicon photonic modulator neuron . Phys. Rev. Appl. ( 2019 ). http://doi.org/10.1103/PhysRevApplied.11.064043 http://doi.org/10.1103/PhysRevApplied.11.064043

B. Mor , S. Garhwal , A. Kumar , A systematic review of hidden Markov models and their applications . Arch. Comput. Methods Eng. 28 , 1429 - 1448 ( 2020 ). http://doi.org/10.1007/s11831-020-09422-4 http://doi.org/10.1007/s11831-020-09422-4

A. Coutrot , J.H. Hsiao , A.B. Chan , Scanpath modeling and classification with hidden Markov models . Behav. Res. Methods 50 , 362 - 379 ( 2017 ). http://doi.org/10.3758/s13428-017-0876-8 http://doi.org/10.3758/s13428-017-0876-8

Y. Khalifa , D. Mandic , E. Sejdić , A review of hidden Markov models and recurrent neural networks for event detection and localization in biomedical signals . Inform. Fusion 69 , 52 - 72 ( 2021 ). http://doi.org/10.1016/j.inffus.2020.11.008 http://doi.org/10.1016/j.inffus.2020.11.008

H. Blanc , B. Dujon , Replicator regions of the yeast mitochondrial DNA responsible for suppressiveness . Proc. Natl. Acad. Sci. 77 , 3942 - 3946 ( 1980 ). http://doi.org/10.1073/pnas.77.7.3942 http://doi.org/10.1073/pnas.77.7.3942

H. Blanc , Two modules from the hypersuppressive rho—mitochondrial DNA are required for plasmid replication in yeast . Gene 30 , 47 - 61 ( 1984 ). http://doi.org/10.1016/0378-1119(84)90104-5 http://doi.org/10.1016/0378-1119(84)90104-5

G.A. Churchill , Stochastic models for heterogeneous DNA sequences . Bull. Math. Biol. 51 , 79 - 94 ( 1989 ). http://doi.org/10.1016/S0092-8240(89)80049-7 http://doi.org/10.1016/S0092-8240(89)80049-7

S.R. Eddy , Hidden Markov models . Curr. Opin. Struct. Biol. 6 , 361 - 365 ( 1996 ). http://doi.org/10.1016/S0959-440X(96)80056-X http://doi.org/10.1016/S0959-440X(96)80056-X

S. Das, A. Tariq, T. Santos, S.S. Kantareddy, I. Banerjee, Machine Learning for Brain Disorders, ed. by Olivier Colliot (Springer, London, 2023), pp. 117–138

A. Lupo , E. Picco , M. Zajnulina , S. Massar , Deep photonic reservoir computer based on frequency multiplexing with fully analog connection between layers . Optica ( 2023 ). http://doi.org/10.1364/optica.489501 http://doi.org/10.1364/optica.489501

T.K. Zhou et al. , Large-scale neuromorphic optoelectronic computing with a reconfigurable diffractive processing unit . Nat. Photonics 15 , 367 - 373 ( 2021 ). http://doi.org/10.1038/s41566-021-00796-w http://doi.org/10.1038/s41566-021-00796-w

B. Wu et al. , Real-valued optical matrix computing with simplified MZI mesh . Intell. Comput. 2 , 0047 ( 2023 ). http://doi.org/10.34133/icomputing.0047 http://doi.org/10.34133/icomputing.0047

M. Kudo, J. Toyama, M. Shimbo, Japanese Vowels Data Set. Distributed by UCI Machine Learning Repository

J. Xu , An extended one-versus-rest support vector machine for multi-label classification . Neurocomputing 74 , 3114 - 3124 ( 2011 ). http://doi.org/10.1016/j.neucom.2011.04.024 http://doi.org/10.1016/j.neucom.2011.04.024

M. Herlihy , N. Shavit , The topological structure of asynchronous computability . J. ACM 46 , 858 - 923 ( 1999 ). http://doi.org/10.1145/331524.331529 http://doi.org/10.1145/331524.331529

H.S. Stokowski et al. , Integrated frequency-modulated optical parametric oscillator . Nature 627 , 95 - 100 ( 2024 ). http://doi.org/10.1038/s41586-024-07071-2 http://doi.org/10.1038/s41586-024-07071-2

A. Jha et al. , Nanophotonic cavity based synapse for scalable photonic neural networks . IEEE J. Sel. Top. Quantum Electron. 28 , 1 - 8 ( 2022 ). http://doi.org/10.1109/jstqe.2022.3179983 http://doi.org/10.1109/jstqe.2022.3179983

A. Rizzo et al. , Massively scalable Kerr comb-driven silicon photonic link . Nat. Photonics ( 2023 ). http://doi.org/10.1038/s41566-023-01244-7 http://doi.org/10.1038/s41566-023-01244-7

J. Zhang , B. Wu , J. Cheng , J. Dong , X. Zhang , Compact, efficient, and scalable nanobeam core for photonic matrix-vector multiplication . Optica ( 2024 ). http://doi.org/10.1364/optica.506603 http://doi.org/10.1364/optica.506603

K. Takada , M. Abe , M. Shibata , M. Ishii , K. Okamoto , Low-crosstalk 10-GHz-spaced 512-channel arrayed-waveguide grating multi/demultiplexer fabricated on a 4-in wafer . IEEE Photonics Technol. Lett. 13 , 1182 - 1184 ( 2001 ). http://doi.org/10.1109/68.959357 http://doi.org/10.1109/68.959357

F. Bohm , G. Verschaffelt , G. Van der Sande , A poor man's coherent Ising machine based on opto-electronic feedback systems for solving optimization problems . Nat. Commun. 10 , 3538 ( 2019 ). http://doi.org/10.1038/s41467-019-11484-3 http://doi.org/10.1038/s41467-019-11484-3

K. Nozaki et al. , Femtofarad optoelectronic integration demonstrating energy-saving signal conversion and nonlinear functions . Nat. Photonics 13 , 454 - 459 ( 2019 ). http://doi.org/10.1038/s41566-019-0397-3 http://doi.org/10.1038/s41566-019-0397-3

Z. Li et al. , Scalable on-chip optoelectronic ising machine utilizing thin-film lithium niobate photonics . ACS Photonics ( 2024 ). http://doi.org/10.1021/acsphotonics.4c00003 http://doi.org/10.1021/acsphotonics.4c00003

Y. Wan et al. , Efficient stochastic parallel gradient descent training for on-chip optical processor . Opto Electron. Adv. 7 , 230182 - 230182 ( 2024 ). http://doi.org/10.29026/oea.2024.230182 http://doi.org/10.29026/oea.2024.230182

D.P. Kingma, J.A. Ba, A method for stochastic optimization. arXiv:1412.6980 https://arxiv.org/abs/1412.6980 (2014).

浏览量

Downloads

CSCD

文章被引用时，请邮件提醒。

Submit

关联资源

暂无数据