TOPS-speed optical tensor convolutional accelerator for feature extraction and inference based on micro-comb

Yixuan Zheng; Shifan Chen; Yifu Xu; Shuai Wang; Zhihui Liu; Yunping Bai; Sai T. Chu; Xiaotian Zhu; Brent E. Little; Roberto Morandotti; David J. Moss; Xingyuan Xu; Kun Xu

doi:10.1364/OFC.2025.Th1C.2

TOPS-speed optical tensor convolutional accelerator for feature extraction and inference based on micro-comb

Yixuan Zheng (Co-first Author)
, Shifan Chen (Co-first Author)
, Yifu Xu
, Shuai Wang
, Zhihui Liu
, Yunping Bai^*
, Sai T. Chu
, Xiaotian Zhu
, Brent E. Little
, Roberto Morandotti
, David J. Moss
, Xingyuan Xu^*
, Kun Xu

^*Corresponding author for this work

Department of Physics

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

Abstract

We experimentally demonstrated an optical tensor convolution accelerator multiplexing the physical dimensions of wavelength, time, and, importantly, space to represent the image channels, achieving an operational speed exceeding 3 TOPS with low data redundancy. © 2025 The Author(s).

Original language	English
Title of host publication	Optical Fiber Communication Conference (OFC) 2025
Publisher	Optica Publishing Group
Number of pages	3
ISBN (Print)	9781557527370
DOIs	https://doi.org/10.1364/OFC.2025.Th1C.2
Publication status	Published - Mar 2025
Event	2025 Optical Fiber Communications Conference and Exhibition (OFC 2025) - San Francisco, United States Duration: 30 Mar 2025 → 3 Apr 2025

Publication series

Name	Optical Fiber Communication Conference in Proceedings Optical Fiber Communication Conference, OFC and Optical Fiber Communication Conference (OFC) Postdeadline Papers

Conference

Conference	2025 Optical Fiber Communications Conference and Exhibition (OFC 2025)
Place	United States
City	San Francisco
Period	30/03/25 → 3/04/25

Funding

This work was supported by the National Natural Science Foundation of China (NSFC) (No,62301074), National Key R&D Program of China (No.2021YFF0901700), NSFC (61821001; 62135009), Fund of State Key Laboratory of IPOC (BUPT) (NO. IPOC2023ZZ01).

Access Output

10.1364/OFC.2025.Th1C.2

Other Access Options

Check CityUHK Library

Permanent Link

Right click to copy link

Cite this

Zheng, Y., Chen, S., Xu, Y., Wang, S., Liu, Z., Bai, Y., Chu, S. T., Zhu, X., Little, B. E., Morandotti, R., Moss, D. J., Xu, X., & Xu, K. (2025). TOPS-speed optical tensor convolutional accelerator for feature extraction and inference based on micro-comb. In Optical Fiber Communication Conference (OFC) 2025 Article Th1C.2 (Optical Fiber Communication Conference in Proceedings Optical Fiber Communication Conference, OFC and Optical Fiber Communication Conference (OFC) Postdeadline Papers). Optica Publishing Group. https://doi.org/10.1364/OFC.2025.Th1C.2

Zheng, Yixuan ; Chen, Shifan ; Xu, Yifu et al. / TOPS-speed optical tensor convolutional accelerator for feature extraction and inference based on micro-comb. Optical Fiber Communication Conference (OFC) 2025. Optica Publishing Group, 2025. (Optical Fiber Communication Conference in Proceedings Optical Fiber Communication Conference, OFC and Optical Fiber Communication Conference (OFC) Postdeadline Papers).

@inproceedings{3635361f35c648df8724e790abe9ffe8,

title = "TOPS-speed optical tensor convolutional accelerator for feature extraction and inference based on micro-comb",

abstract = "We experimentally demonstrated an optical tensor convolution accelerator multiplexing the physical dimensions of wavelength, time, and, importantly, space to represent the image channels, achieving an operational speed exceeding 3 TOPS with low data redundancy. {\textcopyright} 2025 The Author(s).",

author = "Yixuan Zheng and Shifan Chen and Yifu Xu and Shuai Wang and Zhihui Liu and Yunping Bai and Chu, \{Sai T.\} and Xiaotian Zhu and Little, \{Brent E.\} and Roberto Morandotti and Moss, \{David J.\} and Xingyuan Xu and Kun Xu",

year = "2025",

month = mar,

doi = "10.1364/OFC.2025.Th1C.2",

language = "English",

isbn = "9781557527370",

series = "Optical Fiber Communication Conference in Proceedings Optical Fiber Communication Conference, OFC and Optical Fiber Communication Conference (OFC) Postdeadline Papers",

publisher = "Optica Publishing Group",

booktitle = "Optical Fiber Communication Conference (OFC) 2025",

address = "United States",

note = "2025 Optical Fiber Communications Conference and Exhibition (OFC 2025) ; Conference date: 30-03-2025 Through 03-04-2025",

}

Zheng, Y, Chen, S, Xu, Y, Wang, S, Liu, Z, Bai, Y, Chu, ST , Zhu, X, Little, BE, Morandotti, R, Moss, DJ, Xu, X & Xu, K 2025, TOPS-speed optical tensor convolutional accelerator for feature extraction and inference based on micro-comb. in Optical Fiber Communication Conference (OFC) 2025., Th1C.2 , Optical Fiber Communication Conference in Proceedings Optical Fiber Communication Conference, OFC and Optical Fiber Communication Conference (OFC) Postdeadline Papers, Optica Publishing Group, 2025 Optical Fiber Communications Conference and Exhibition (OFC 2025), San Francisco, California, United States, 30/03/25. https://doi.org/10.1364/OFC.2025.Th1C.2

TOPS-speed optical tensor convolutional accelerator for feature extraction and inference based on micro-comb. / Zheng, Yixuan (Co-first Author); Chen, Shifan (Co-first Author); Xu, Yifu et al.
Optical Fiber Communication Conference (OFC) 2025. Optica Publishing Group, 2025. Th1C.2 (Optical Fiber Communication Conference in Proceedings Optical Fiber Communication Conference, OFC and Optical Fiber Communication Conference (OFC) Postdeadline Papers).

Research output: Chapters, Conference Papers, Creative and Literary Works › RGC 32 - Refereed conference paper (with host publication) › peer-review

TY - GEN

T1 - TOPS-speed optical tensor convolutional accelerator for feature extraction and inference based on micro-comb

AU - Zheng, Yixuan

AU - Chen, Shifan

AU - Xu, Yifu

AU - Wang, Shuai

AU - Liu, Zhihui

AU - Bai, Yunping

AU - Chu, Sai T.

AU - Zhu, Xiaotian

AU - Little, Brent E.

AU - Morandotti, Roberto

AU - Moss, David J.

AU - Xu, Xingyuan

AU - Xu, Kun

PY - 2025/3

Y1 - 2025/3

N2 - We experimentally demonstrated an optical tensor convolution accelerator multiplexing the physical dimensions of wavelength, time, and, importantly, space to represent the image channels, achieving an operational speed exceeding 3 TOPS with low data redundancy. © 2025 The Author(s).

AB - We experimentally demonstrated an optical tensor convolution accelerator multiplexing the physical dimensions of wavelength, time, and, importantly, space to represent the image channels, achieving an operational speed exceeding 3 TOPS with low data redundancy. © 2025 The Author(s).

UR - https://www.scopus.com/pages/publications/105011356277

UR - https://www.scopus.com/record/pubmetrics.uri?eid=2-s2.0-105011356277&origin=recordpage

U2 - 10.1364/OFC.2025.Th1C.2

DO - 10.1364/OFC.2025.Th1C.2

M3 - RGC 32 - Refereed conference paper (with host publication)

SN - 9781557527370

T3 - Optical Fiber Communication Conference in Proceedings Optical Fiber Communication Conference, OFC and Optical Fiber Communication Conference (OFC) Postdeadline Papers

BT - Optical Fiber Communication Conference (OFC) 2025

PB - Optica Publishing Group

T2 - 2025 Optical Fiber Communications Conference and Exhibition (OFC 2025)

Y2 - 30 March 2025 through 3 April 2025

ER -

Zheng Y, Chen S, Xu Y, Wang S, Liu Z, Bai Y et al. TOPS-speed optical tensor convolutional accelerator for feature extraction and inference based on micro-comb. In Optical Fiber Communication Conference (OFC) 2025. Optica Publishing Group. 2025. Th1C.2 . (Optical Fiber Communication Conference in Proceedings Optical Fiber Communication Conference, OFC and Optical Fiber Communication Conference (OFC) Postdeadline Papers). doi: 10.1364/OFC.2025.Th1C.2

TOPS-speed optical tensor convolutional accelerator for feature extraction and inference based on micro-comb

Abstract

Publication series

Conference

Funding

Access Output

Other Access Options

Permanent Link

Other Files and Links

Fingerprint

Cite this