Cargando…

Meta-optic accelerators for object classifiers

Rapid advances in deep learning have led to paradigm shifts in a number of fields, from medical image analysis to autonomous systems. These advances, however, have resulted in digital neural networks with large computational requirements, resulting in high energy consumption and limitations in real-...

Descripción completa

Detalles Bibliográficos
Autores principales: Zheng, Hanyu, Liu, Quan, Zhou, You, Kravchenko, Ivan I., Huo, Yuankai, Valentine, Jason
Formato: Online Artículo Texto
Lenguaje:English
Publicado: American Association for the Advancement of Science 2022
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9328681/
https://www.ncbi.nlm.nih.gov/pubmed/35895828
http://dx.doi.org/10.1126/sciadv.abo6410
_version_ 1784757777486315520
author Zheng, Hanyu
Liu, Quan
Zhou, You
Kravchenko, Ivan I.
Huo, Yuankai
Valentine, Jason
author_facet Zheng, Hanyu
Liu, Quan
Zhou, You
Kravchenko, Ivan I.
Huo, Yuankai
Valentine, Jason
author_sort Zheng, Hanyu
collection PubMed
description Rapid advances in deep learning have led to paradigm shifts in a number of fields, from medical image analysis to autonomous systems. These advances, however, have resulted in digital neural networks with large computational requirements, resulting in high energy consumption and limitations in real-time decision-making when computation resources are limited. Here, we demonstrate a meta-optic–based neural network accelerator that can off-load computationally expensive convolution operations into high-speed and low-power optics. In this architecture, metasurfaces enable both spatial multiplexing and additional information channels, such as polarization, in object classification. End-to-end design is used to co-optimize the optical and digital systems, resulting in a robust classifier that achieves 93.1% accurate classification of handwriting digits and 93.8% accuracy in classifying both the digit and its polarization state. This approach could enable compact, high-speed, and low-power image and information processing systems for a wide range of applications in machine vision and artificial intelligence.
format Online
Article
Text
id pubmed-9328681
institution National Center for Biotechnology Information
language English
publishDate 2022
publisher American Association for the Advancement of Science
record_format MEDLINE/PubMed
spelling pubmed-93286812022-08-09 Meta-optic accelerators for object classifiers Zheng, Hanyu Liu, Quan Zhou, You Kravchenko, Ivan I. Huo, Yuankai Valentine, Jason Sci Adv Physical and Materials Sciences Rapid advances in deep learning have led to paradigm shifts in a number of fields, from medical image analysis to autonomous systems. These advances, however, have resulted in digital neural networks with large computational requirements, resulting in high energy consumption and limitations in real-time decision-making when computation resources are limited. Here, we demonstrate a meta-optic–based neural network accelerator that can off-load computationally expensive convolution operations into high-speed and low-power optics. In this architecture, metasurfaces enable both spatial multiplexing and additional information channels, such as polarization, in object classification. End-to-end design is used to co-optimize the optical and digital systems, resulting in a robust classifier that achieves 93.1% accurate classification of handwriting digits and 93.8% accuracy in classifying both the digit and its polarization state. This approach could enable compact, high-speed, and low-power image and information processing systems for a wide range of applications in machine vision and artificial intelligence. American Association for the Advancement of Science 2022-07-27 /pmc/articles/PMC9328681/ /pubmed/35895828 http://dx.doi.org/10.1126/sciadv.abo6410 Text en Copyright © 2022 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim to original U.S. Government Works. Distributed under a Creative Commons Attribution License 4.0 (CC BY). https://creativecommons.org/licenses/by/4.0/This is an open-access article distributed under the terms of the Creative Commons Attribution license (https://creativecommons.org/licenses/by/4.0/) , which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
spellingShingle Physical and Materials Sciences
Zheng, Hanyu
Liu, Quan
Zhou, You
Kravchenko, Ivan I.
Huo, Yuankai
Valentine, Jason
Meta-optic accelerators for object classifiers
title Meta-optic accelerators for object classifiers
title_full Meta-optic accelerators for object classifiers
title_fullStr Meta-optic accelerators for object classifiers
title_full_unstemmed Meta-optic accelerators for object classifiers
title_short Meta-optic accelerators for object classifiers
title_sort meta-optic accelerators for object classifiers
topic Physical and Materials Sciences
url https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9328681/
https://www.ncbi.nlm.nih.gov/pubmed/35895828
http://dx.doi.org/10.1126/sciadv.abo6410
work_keys_str_mv AT zhenghanyu metaopticacceleratorsforobjectclassifiers
AT liuquan metaopticacceleratorsforobjectclassifiers
AT zhouyou metaopticacceleratorsforobjectclassifiers
AT kravchenkoivani metaopticacceleratorsforobjectclassifiers
AT huoyuankai metaopticacceleratorsforobjectclassifiers
AT valentinejason metaopticacceleratorsforobjectclassifiers