Modeling of Optical Matrix Multipliers Using Transposed Convolutional Neural Networks

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 wi...

Full description

Saved in:
Bibliographic Details
Published in2022 IEEE Photonics Conference (IPC) pp. 1 - 2
Main Authors Cem, Ali, Yan, Siqi, De Moura, Uiara Celine, Ding, Yunhong, Zibar, Darko, Da Ros, Francesco
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.11.2022
Subjects
machine learning
neuromorphic photonics
optical matrix multiplication
transposed convolutional neural network
Online AccessGet full text

Cover

More Information
Summary: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).
ISSN:2575-274X
DOI:10.1109/IPC53466.2022.9975676