Scalable physical deep learning using optical dynamics with state-skipping training

Scalable physical deep learning using optical dynamics with state-skipping training | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Scalable physical deep learning using optical dynamics with state-skipping training Yongbo Zhang, Mitsumasa Nakajima, Katsuma Inoue, Toshikazu Hashimoto, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8145052/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted You are reading this latest preprint version Abstract The escalating energy demands for high-performance machine learning have sparked growing interest in unconventional computing paradigms rooted in physical systems. At the core of this emerging direction is the leveraging of physical properties to design computing frameworks that integrate strong learning capabilities, efficient training strategies, and practical implementability. Here, we present a simple, efficient, and scalable framework aimed at achieving this, and validate it on an optoelectronic platform. This framework features our proposed training mechanism—state-skipping direct feedback alignment—which eliminates access to intermediate states of the system and thereby significantly simplifies the error backpropagation process, enhancing both training efficiency and practical feasibility. Compared to conventional deep neural networks, our approach substantially reduces computational costs and training parameter count while achieving comparable performance. Furthermore, we integrate our scheme into modern architectures, attaining improved performance with an approximately 40% reduction in computational resources. Notably, our approach challenges the scaling laws of conventional counterparts, underscoring its strong scalability and practical promise. Physical sciences/Mathematics and computing/Information technology Physical sciences/Mathematics and computing/Computational science Physical sciences/Optics and photonics/Applied optics/Fibre optics and optical communications Full Text Additional Declarations There is NO Competing Interest. Supplementary Files Supplementaryinformation.pdf Supplementary Information Cite Share Download PDF Status: Under Review Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8145052","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":573235992,"identity":"91b63ebc-3a53-43e1-af8c-dca5e09623c4","order_by":0,"name":"Yongbo Zhang","email":"data:image/png;base64,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","orcid":"https://orcid.org/0009-0003-8475-2145","institution":"The University of Tokyo","correspondingAuthor":true,"prefix":"","firstName":"Yongbo","middleName":"","lastName":"Zhang","suffix":""},{"id":573235993,"identity":"9f507933-740c-469f-95fd-fb13bd78fe0b","order_by":1,"name":"Mitsumasa Nakajima","email":"","orcid":"","institution":"NTT Device Technology Laboratories","correspondingAuthor":false,"prefix":"","firstName":"Mitsumasa","middleName":"","lastName":"Nakajima","suffix":""},{"id":573235994,"identity":"b4a38825-8bba-4766-92c8-fe401009e703","order_by":2,"name":"Katsuma Inoue","email":"","orcid":"","institution":"The University of Tokyo","correspondingAuthor":false,"prefix":"","firstName":"Katsuma","middleName":"","lastName":"Inoue","suffix":""},{"id":573235995,"identity":"e09790a6-39f3-49d3-94cf-3a7e0d55f834","order_by":3,"name":"Toshikazu Hashimoto","email":"","orcid":"","institution":"NTT Device Technology Laboratories","correspondingAuthor":false,"prefix":"","firstName":"Toshikazu","middleName":"","lastName":"Hashimoto","suffix":""},{"id":573235996,"identity":"cd3ccb70-b67a-401a-81ce-d6c481f3fc07","order_by":4,"name":"Yasuo Kuniyoshi","email":"","orcid":"https://orcid.org/0000-0001-8443-4161","institution":"Graduate School of Information Science \u0026 Technology, The University of Tokyo","correspondingAuthor":false,"prefix":"","firstName":"Yasuo","middleName":"","lastName":"Kuniyoshi","suffix":""},{"id":573235997,"identity":"58f7a1a3-5d34-4d47-afbe-78f18974a830","order_by":5,"name":"Kohei Nakajima","email":"","orcid":"https://orcid.org/0000-0001-5589-4054","institution":"The University of Tokyo","correspondingAuthor":false,"prefix":"","firstName":"Kohei","middleName":"","lastName":"Nakajima","suffix":""}],"badges":[],"createdAt":"2025-11-18 11:55:25","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8145052/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8145052/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":100366044,"identity":"02db4c19-3d2a-4a1e-a37b-e7889cc9d5a9","added_by":"auto","created_at":"2026-01-16 07:55:53","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":3178875,"visible":true,"origin":"","legend":"Article File","description":"","filename":"Maintext.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8145052/v1_covered_3968542d-f4a6-4897-8991-3e5cd0bf500d.pdf"},{"id":100130092,"identity":"3940f3b1-8061-4976-b248-5f6c4353c520","added_by":"auto","created_at":"2026-01-13 10:03:24","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2650361,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"Supplementaryinformation.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8145052/v1/af3c92a1e831766b1cfce321.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Scalable physical deep learning using optical dynamics with state-skipping training","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"nature-portfolio","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"","title":"Nature Portfolio","twitterHandle":"","acdcEnabled":false,"dfaEnabled":false,"editorialSystem":"ejp","reportingPortfolio":"","inReviewEnabled":true,"inReviewRevisionsEnabled":false},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-8145052/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8145052/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The escalating energy demands for high-performance machine learning have sparked growing interest in unconventional computing paradigms rooted in physical systems. At the core of this emerging direction is the leveraging of physical properties to design computing frameworks that integrate strong learning capabilities, efficient training strategies, and practical implementability. Here, we present a simple, efficient, and scalable framework aimed at achieving this, and validate it on an optoelectronic platform. This framework features our proposed training mechanism—state-skipping direct feedback alignment—which eliminates access to intermediate states of the system and thereby significantly simplifies the error backpropagation process, enhancing both training efficiency and practical feasibility. Compared to conventional deep neural networks, our approach substantially reduces computational costs and training parameter count while achieving comparable performance. Furthermore, we integrate our scheme into modern architectures, attaining improved performance with an approximately 40% reduction in computational resources. Notably, our approach challenges the scaling laws of conventional counterparts, underscoring its strong scalability and practical promise.","manuscriptTitle":"Scalable physical deep learning using optical dynamics with state-skipping training","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-01-13 10:03:19","doi":"10.21203/rs.3.rs-8145052/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-computational-science","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"natcomputsci","sideBox":"Learn more about [Nature Computational Science](http://www.nature.com/natcomputsci/)","snPcode":"","submissionUrl":"","title":"Nature Computational Science","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Research","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"1ca370e8-a3ec-47ba-b3af-a0370407975f","owner":[],"postedDate":"January 13th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":60989849,"name":"Physical sciences/Mathematics and computing/Information technology"},{"id":60989850,"name":"Physical sciences/Mathematics and computing/Computational science"},{"id":60989851,"name":"Physical sciences/Optics and photonics/Applied optics/Fibre optics and optical communications"}],"tags":[],"updatedAt":"2026-04-24T17:45:22+00:00","versionOfRecord":[],"versionCreatedAt":"2026-01-13 10:03:19","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8145052","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8145052","identity":"rs-8145052","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}