扩展端到端片上光子神经网络推理。

IF 23.4 Q1 OPTICS
Bo Wu,Chaoran Huang,Jialong Zhang,Hailong Zhou,Yilun Wang,Jianji Dong,Xinliang Zhang
{"title":"扩展端到端片上光子神经网络推理。","authors":"Bo Wu,Chaoran Huang,Jialong Zhang,Hailong Zhou,Yilun Wang,Jianji Dong,Xinliang Zhang","doi":"10.1038/s41377-025-02029-z","DOIUrl":null,"url":null,"abstract":"Optical neural networks are emerging as a competitive alternative to their electronic counterparts, offering distinct advantages in bandwidth and energy efficiency. Despite these benefits, scaling up on-chip optical neural networks for end-to-end inference is facing significant challenges. First, network depth is constrained by the weak cascadability of optical nonlinear activation functions. Second, the input size is constrained by the scale of the optical matrix. Herein, we propose a scaling up strategy called partially coherent deep optical neural networks (PDONNs). By leveraging an on-chip nonlinear activation function based on opto-electro-opto conversion, PDONN enables network depth expansion with positive net gain. Additionally, convolutional layers achieve rapid dimensionality reduction, thereby allowing for an increase in the accommodated input size. The use of a partially coherent optical source significantly reduces reliance on narrow-linewidth laser diodes and coherent detection. Owing to their broader spectral characteristics and simpler implementation, such sources are more accessible and compatible with scalable integration. Benefiting from these innovations, we designed and fabricated a monolithically integrated optical neural network with the largest input size and the deepest network depth, comprising an input layer with a size of 64, two convolutional layers, and two fully connected layers. We successfully demonstrate end-to-end two-class classification of fashion images and four-class classification of handwritten digits with accuracies of 96% and 94%, respectively, using an in-situ training method. Notably, performance is well maintained with partially coherent illumination. This proposed architecture represents a critical step toward realizing energy-efficient, scalable, and widely accessible optical computing.","PeriodicalId":18069,"journal":{"name":"Light-Science & Applications","volume":"4 1","pages":"328"},"PeriodicalIF":23.4000,"publicationDate":"2025-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Scaling up for end-to-end on-chip photonic neural network inference.\",\"authors\":\"Bo Wu,Chaoran Huang,Jialong Zhang,Hailong Zhou,Yilun Wang,Jianji Dong,Xinliang Zhang\",\"doi\":\"10.1038/s41377-025-02029-z\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Optical neural networks are emerging as a competitive alternative to their electronic counterparts, offering distinct advantages in bandwidth and energy efficiency. Despite these benefits, scaling up on-chip optical neural networks for end-to-end inference is facing significant challenges. First, network depth is constrained by the weak cascadability of optical nonlinear activation functions. Second, the input size is constrained by the scale of the optical matrix. Herein, we propose a scaling up strategy called partially coherent deep optical neural networks (PDONNs). By leveraging an on-chip nonlinear activation function based on opto-electro-opto conversion, PDONN enables network depth expansion with positive net gain. Additionally, convolutional layers achieve rapid dimensionality reduction, thereby allowing for an increase in the accommodated input size. The use of a partially coherent optical source significantly reduces reliance on narrow-linewidth laser diodes and coherent detection. Owing to their broader spectral characteristics and simpler implementation, such sources are more accessible and compatible with scalable integration. Benefiting from these innovations, we designed and fabricated a monolithically integrated optical neural network with the largest input size and the deepest network depth, comprising an input layer with a size of 64, two convolutional layers, and two fully connected layers. We successfully demonstrate end-to-end two-class classification of fashion images and four-class classification of handwritten digits with accuracies of 96% and 94%, respectively, using an in-situ training method. Notably, performance is well maintained with partially coherent illumination. This proposed architecture represents a critical step toward realizing energy-efficient, scalable, and widely accessible optical computing.\",\"PeriodicalId\":18069,\"journal\":{\"name\":\"Light-Science & Applications\",\"volume\":\"4 1\",\"pages\":\"328\"},\"PeriodicalIF\":23.4000,\"publicationDate\":\"2025-09-17\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Light-Science & Applications\",\"FirstCategoryId\":\"1089\",\"ListUrlMain\":\"https://doi.org/10.1038/s41377-025-02029-z\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q1\",\"JCRName\":\"OPTICS\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Light-Science & Applications","FirstCategoryId":"1089","ListUrlMain":"https://doi.org/10.1038/s41377-025-02029-z","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"OPTICS","Score":null,"Total":0}
引用次数: 0

摘要

光神经网络在带宽和能源效率方面具有明显的优势,正在成为电子网络的竞争替代品。尽管有这些好处,扩展端到端推理的片上光学神经网络仍面临着重大挑战。首先,网络深度受到光学非线性激活函数的弱级联性的限制。其次,输入尺寸受光学矩阵尺度的限制。在此,我们提出了一种称为部分相干深度光神经网络(PDONNs)的扩展策略。通过利用基于光电转换的片上非线性激活函数,PDONN使网络深度扩展具有正净增益。此外,卷积层实现了快速降维,从而允许增加可容纳的输入大小。部分相干光源的使用大大减少了对窄线宽激光二极管和相干检测的依赖。由于其更广泛的光谱特性和更简单的实现,这些源更容易获得,并且与可扩展集成兼容。得益于这些创新,我们设计并制造了一个具有最大输入尺寸和最深网络深度的单片集成光学神经网络,包括大小为64的输入层,两个卷积层和两个完全连接层。我们使用原位训练方法成功地演示了端到端时尚图像的两类分类和手写数字的四类分类,准确率分别为96%和94%。值得注意的是,在部分相干照明下,性能保持得很好。提出的体系结构是实现节能、可扩展和可广泛访问的光学计算的关键一步。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Scaling up for end-to-end on-chip photonic neural network inference.
Optical neural networks are emerging as a competitive alternative to their electronic counterparts, offering distinct advantages in bandwidth and energy efficiency. Despite these benefits, scaling up on-chip optical neural networks for end-to-end inference is facing significant challenges. First, network depth is constrained by the weak cascadability of optical nonlinear activation functions. Second, the input size is constrained by the scale of the optical matrix. Herein, we propose a scaling up strategy called partially coherent deep optical neural networks (PDONNs). By leveraging an on-chip nonlinear activation function based on opto-electro-opto conversion, PDONN enables network depth expansion with positive net gain. Additionally, convolutional layers achieve rapid dimensionality reduction, thereby allowing for an increase in the accommodated input size. The use of a partially coherent optical source significantly reduces reliance on narrow-linewidth laser diodes and coherent detection. Owing to their broader spectral characteristics and simpler implementation, such sources are more accessible and compatible with scalable integration. Benefiting from these innovations, we designed and fabricated a monolithically integrated optical neural network with the largest input size and the deepest network depth, comprising an input layer with a size of 64, two convolutional layers, and two fully connected layers. We successfully demonstrate end-to-end two-class classification of fashion images and four-class classification of handwritten digits with accuracies of 96% and 94%, respectively, using an in-situ training method. Notably, performance is well maintained with partially coherent illumination. This proposed architecture represents a critical step toward realizing energy-efficient, scalable, and widely accessible optical computing.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
Light-Science & Applications
Light-Science & Applications 数理科学, 物理学I, 光学, 凝聚态物性 II :电子结构、电学、磁学和光学性质, 无机非金属材料, 无机非金属类光电信息与功能材料, 工程与材料, 信息科学, 光学和光电子学, 光学和光电子材料, 非线性光学与量子光学
自引率
0.00%
发文量
803
审稿时长
2.1 months
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信