Modeling of Optical Matrix Multipliers Using Transposed Convolutional Neural Networks

Ali Cem; Siqi Yan; Uiara Celine De Moura; Yunhong Ding; Darko Zibar; Francesco Da Ros

doi:10.1109/IPC53466.2022.9975676

Modeling of Optical Matrix Multipliers Using Transposed Convolutional Neural Networks

Ali Cem
, Siqi Yan
, Uiara Celine De Moura
, Yunhong Ding
, Darko Zibar
, Francesco Da Ros

Huazhong University of Science and Technology

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

145 Downloads (Orbit)

Abstract

We demonstrate a data-driven model for optical matrix multipliers utilizing Mach-Zehnder interferometer meshes. For a fabricated chip, a transposed convolutional neural network model learns from experimental measurements offline and predicts the weights across 100 frequency channels in the C-band with high precision (RMSE<0.8 dB).

Original language	English
Title of host publication	Proceedings of 2022 IEEE Photonics Conference
Number of pages	2
Publisher	IEEE
Publication date	17 Nov 2022
Article number	9975676
ISBN (Print)	978-1-6654-3488-1
DOIs	https://doi.org/10.1109/IPC53466.2022.9975676
Publication status	Published - 17 Nov 2022
Event	2022 IEEE Photonics Conference - Hyatt Regency Vancouver Hotel, Vancouver, Canada Duration: 13 Nov 2022 → 17 Nov 2022

Conference

Conference	2022 IEEE Photonics Conference
Location	Hyatt Regency Vancouver Hotel
Country/Territory	Canada
City	Vancouver
Period	13/11/2022 → 17/11/2022

Keywords

Weight measurement
Integrated optics
Optical interferometry
Optical device fabrication
Optical computing
Predictive models
Adaptive optics

Access to Document

10.1109/IPC53466.2022.9975676

FulltextAccepted author manuscript, 337 KB

OpenUrl availability

Full text

Cite this

@inproceedings{024e4949967746a9b5791c75d93b7f17,

title = "Modeling of Optical Matrix Multipliers Using Transposed Convolutional Neural Networks",

abstract = "We demonstrate a data-driven model for optical matrix multipliers utilizing Mach-Zehnder interferometer meshes. For a fabricated chip, a transposed convolutional neural network model learns from experimental measurements offline and predicts the weights across 100 frequency channels in the C-band with high precision (RMSE<0.8 dB).",

keywords = "Weight measurement, Integrated optics, Optical interferometry, Optical device fabrication, Optical computing, Predictive models, Adaptive optics",

author = "Ali Cem and Siqi Yan and Moura, \{Uiara Celine De\} and Yunhong Ding and Darko Zibar and Ros, \{Francesco Da\}",

year = "2022",

month = nov,

day = "17",

doi = "10.1109/IPC53466.2022.9975676",

language = "English",

isbn = "978-1-6654-3488-1",

booktitle = "Proceedings of 2022 IEEE Photonics Conference",

publisher = "IEEE",

address = "United States",

note = "2022 IEEE Photonics Conference, IPC 2022 ; Conference date: 13-11-2022 Through 17-11-2022",

}

TY - GEN

T1 - Modeling of Optical Matrix Multipliers Using Transposed Convolutional Neural Networks

AU - Cem, Ali

AU - Yan, Siqi

AU - Moura, Uiara Celine De

AU - Ding, Yunhong

AU - Zibar, Darko

AU - Ros, Francesco Da

PY - 2022/11/17

Y1 - 2022/11/17

N2 - We demonstrate a data-driven model for optical matrix multipliers utilizing Mach-Zehnder interferometer meshes. For a fabricated chip, a transposed convolutional neural network model learns from experimental measurements offline and predicts the weights across 100 frequency channels in the C-band with high precision (RMSE<0.8 dB).

AB - We demonstrate a data-driven model for optical matrix multipliers utilizing Mach-Zehnder interferometer meshes. For a fabricated chip, a transposed convolutional neural network model learns from experimental measurements offline and predicts the weights across 100 frequency channels in the C-band with high precision (RMSE<0.8 dB).

KW - Weight measurement

KW - Integrated optics

KW - Optical interferometry

KW - Optical device fabrication

KW - Optical computing

KW - Predictive models

KW - Adaptive optics

U2 - 10.1109/IPC53466.2022.9975676

DO - 10.1109/IPC53466.2022.9975676

M3 - Article in proceedings

SN - 978-1-6654-3488-1

BT - Proceedings of 2022 IEEE Photonics Conference

PB - IEEE

T2 - 2022 IEEE Photonics Conference

Y2 - 13 November 2022 through 17 November 2022

ER -

Modeling of Optical Matrix Multipliers Using Transposed Convolutional Neural Networks

Abstract

Conference

Keywords

Access to Document

OpenUrl availability

Fingerprint

Cite this