{"title":"Analogue optical pattern recognition for cross-correlational CNN.","authors":"Ahmed Farhat, Wim J C Melis","doi":"10.1111/jmi.70034","DOIUrl":null,"url":null,"abstract":"<p><p>Pattern recognition in convolutional neural networks (CNNs) is computationally intensive due to its reliance on 2D convolutions, requiring significant processing power and time. This paper proposes an analogue optical hardware system to improve CNN efficiency, focusing on forward propagation tasks such as data preparation, correlation, and decision-making. By utilising the continuous properties of light waves for 2D convolutional operations, the system overcomes key limitations of von Neumann architectures around saving power and time. Optical wave operations allow for more efficient and instantaneous tasks like 2D Fourier transforms, which are crucial to pattern recognition. The paper validates these concepts through simulations using MATLAB and COMSOL. Overall, the presented approach paves the way for more efficient ML hardware. Future work will focus on extending the system to enable full CNN training, including backward propagation, as well as the development of commercially suitable hardware implementations.</p>","PeriodicalId":16484,"journal":{"name":"Journal of microscopy","volume":" ","pages":""},"PeriodicalIF":1.9000,"publicationDate":"2025-09-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of microscopy","FirstCategoryId":"5","ListUrlMain":"https://doi.org/10.1111/jmi.70034","RegionNum":4,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"MICROSCOPY","Score":null,"Total":0}
引用次数: 0
Abstract
Pattern recognition in convolutional neural networks (CNNs) is computationally intensive due to its reliance on 2D convolutions, requiring significant processing power and time. This paper proposes an analogue optical hardware system to improve CNN efficiency, focusing on forward propagation tasks such as data preparation, correlation, and decision-making. By utilising the continuous properties of light waves for 2D convolutional operations, the system overcomes key limitations of von Neumann architectures around saving power and time. Optical wave operations allow for more efficient and instantaneous tasks like 2D Fourier transforms, which are crucial to pattern recognition. The paper validates these concepts through simulations using MATLAB and COMSOL. Overall, the presented approach paves the way for more efficient ML hardware. Future work will focus on extending the system to enable full CNN training, including backward propagation, as well as the development of commercially suitable hardware implementations.
期刊介绍:
The Journal of Microscopy is the oldest journal dedicated to the science of microscopy and the only peer-reviewed publication of the Royal Microscopical Society. It publishes papers that report on the very latest developments in microscopy such as advances in microscopy techniques or novel areas of application. The Journal does not seek to publish routine applications of microscopy or specimen preparation even though the submission may otherwise have a high scientific merit.
The scope covers research in the physical and biological sciences and covers imaging methods using light, electrons, X-rays and other radiations as well as atomic force and near field techniques. Interdisciplinary research is welcome. Papers pertaining to microscopy are also welcomed on optical theory, spectroscopy, novel specimen preparation and manipulation methods and image recording, processing and analysis including dynamic analysis of living specimens.
Publication types include full papers, hot topic fast tracked communications and review articles. Authors considering submitting a review article should contact the editorial office first.