Optical Neuromorphic Processor at 11 TeraOPs/s based on Kerr Soliton Crystal Micro-combs

Mengxi Tan; Xingyuan Xu; Jiayang Wu; Andreas Boes; Bill Corcoran; Thach G. Nguyen; Sai T. Chu; Brent E. Little; Damien G. Hicks; Roberto Morandotti; Arnan Mitchell; David J. Moss

Optical Neuromorphic Processor at 11 TeraOPs/s based on Kerr Soliton Crystal Micro-combs

Mengxi Tan
, Xingyuan Xu
, Jiayang Wu
, Andreas Boes
, Bill Corcoran
, Thach G. Nguyen
, Sai T. Chu
, Brent E. Little
, Damien G. Hicks
, Roberto Morandotti
, Arnan Mitchell
, David J. Moss^*

^*Corresponding author for this work

Research output: Contribution to journal › Conference article › peer-review

1 Scopus citations

Abstract

We demonstrate a universal optical vector convolutional accelerator operating at 11 Tera-OPS, generating convolutions of images of 250,000 pixels with 8-bit resolution for 10 kernels simultaneously. We use the same hardware to form a deep optical CNN with ten output neurons, achieving successful recognition of full 10 digits with 88% accuracy. Our approach is scalable and trainable for applications to unmanned vehicle and real-time video recognition.

Original language	English
Article number	Tu3G.1
Journal	Optics InfoBase Conference Papers
State	Published - 2022
Externally published	Yes
Event	Optical Fiber Communication Conference, OFC 2022 - San Diego, United States Duration: 6 Mar 2022 → 10 Mar 2022

Cite this

@article{ca17612efdff4603a68aadc358074290,

title = "Optical Neuromorphic Processor at 11 TeraOPs/s based on Kerr Soliton Crystal Micro-combs",

abstract = "We demonstrate a universal optical vector convolutional accelerator operating at 11 Tera-OPS, generating convolutions of images of 250,000 pixels with 8-bit resolution for 10 kernels simultaneously. We use the same hardware to form a deep optical CNN with ten output neurons, achieving successful recognition of full 10 digits with 88\% accuracy. Our approach is scalable and trainable for applications to unmanned vehicle and real-time video recognition.",

author = "Mengxi Tan and Xingyuan Xu and Jiayang Wu and Andreas Boes and Bill Corcoran and Nguyen, \{Thach G.\} and Chu, \{Sai T.\} and Little, \{Brent E.\} and Hicks, \{Damien G.\} and Roberto Morandotti and Arnan Mitchell and Moss, \{David J.\}",

note = "Publisher Copyright: {\textcopyright} Optica Publishing Group 2022, {\textcopyright} 2022 The Author(s); Optical Fiber Communication Conference, OFC 2022 ; Conference date: 06-03-2022 Through 10-03-2022",

year = "2022",

language = "英语",

journal = "Optics InfoBase Conference Papers",

issn = "2162-2701",

}

TY - JOUR

T1 - Optical Neuromorphic Processor at 11 TeraOPs/s based on Kerr Soliton Crystal Micro-combs

AU - Tan, Mengxi

AU - Xu, Xingyuan

AU - Wu, Jiayang

AU - Boes, Andreas

AU - Corcoran, Bill

AU - Nguyen, Thach G.

AU - Chu, Sai T.

AU - Little, Brent E.

AU - Hicks, Damien G.

AU - Morandotti, Roberto

AU - Mitchell, Arnan

AU - Moss, David J.

PY - 2022

Y1 - 2022

N2 - We demonstrate a universal optical vector convolutional accelerator operating at 11 Tera-OPS, generating convolutions of images of 250,000 pixels with 8-bit resolution for 10 kernels simultaneously. We use the same hardware to form a deep optical CNN with ten output neurons, achieving successful recognition of full 10 digits with 88% accuracy. Our approach is scalable and trainable for applications to unmanned vehicle and real-time video recognition.

AB - We demonstrate a universal optical vector convolutional accelerator operating at 11 Tera-OPS, generating convolutions of images of 250,000 pixels with 8-bit resolution for 10 kernels simultaneously. We use the same hardware to form a deep optical CNN with ten output neurons, achieving successful recognition of full 10 digits with 88% accuracy. Our approach is scalable and trainable for applications to unmanned vehicle and real-time video recognition.

UR - https://www.scopus.com/pages/publications/85168058966

M3 - 会议文章

AN - SCOPUS:85168058966

SN - 2162-2701

JO - Optics InfoBase Conference Papers

JF - Optics InfoBase Conference Papers

M1 - Tu3G.1

T2 - Optical Fiber Communication Conference, OFC 2022

Y2 - 6 March 2022 through 10 March 2022

ER -

Optical Neuromorphic Processor at 11 TeraOPs/s based on Kerr Soliton Crystal Micro-combs

Abstract

Other files and links

Fingerprint

Cite this