SOT-MRAM-enabled noise-tolerant and resource-saving probabilistic binary neural network

SOT-MRAM-enabled noise-tolerant and resource-saving probabilistic binary neural network | 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 Physical Sciences - Article SOT-MRAM-enabled noise-tolerant and resource-saving probabilistic binary neural network Xufeng Kou, Puyang Huang, Yu Gu, Chenyi Fu, Tianhao Chen, Shan Yao, and 10 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5905772/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 deployment of large-scale full-precision neural networks is hindered by the escalating parameter sizes and inherent susceptibility to noise. Here, we present a non-volatile memory (NVM)-based probabilistic computing architecture, which harnesses both the stochastic nature and in-memory computing capabilities of NVM to enhance computation efficiency and robustness. Utilizing in-plane magnetized spin-orbit torque magneto-resistive random-access memory (i-SOT MRAM) cells, we demonstrate ultra-fast (400 ps), field-free write operations and voltage-controllable probabilistic states with a low variation of 3.8%. These features enable the implementation of an on-chip SOT probabilistic binary neural network (SOT-PBNN) hardware, achieving near-baseline accuracy (88.2%) on CIFAR-10 classification tasks. Moreover, the stochastic training process endows the SOT-PBNN to attain a 7-fold improvement in classification accuracy over full-precision networks under a 25% write/read noise, while reducing the parameter size by 1–2 orders of magnitude. Our work establishes a lightweight framework suitable for realizing artificial intelligence platforms on resource-constrained edge devices. Physical sciences/Materials science/Condensed-matter physics/Magnetic properties and materials Physical sciences/Materials science/Condensed-matter physics/Ferromagnetism Full Text Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryInformation.docx 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-5905772","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Physical Sciences - Article","associatedPublications":[],"authors":[{"id":414154877,"identity":"2536e791-af0e-42f0-9157-5cf306fb6a54","order_by":0,"name":"Xufeng Kou","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0klEQVRIiWNgGAWjYBACCQbGhw8/VNTIgTgHHhCnhdnYWOLMMWOwlgQitZhJ8LYxJzaAeERpkZyRzCAhwcaWPj/s8EOgLXZyug0EtEjzHGYwKOCRyd14O80AqCXZ2OwAAS1y7P0HEoDW5G6cnQDSciBxG0EtzMwMB3gMmNMNZ6d/IE6LNHszYwNPAnOCvHQOkbZI9hxmZpY4cMxwg3ROwYEEAyL8InEjmf3nx3818vKz0zd/+FBhJ0dQCxwYgFUaEKscBOQbSFE9CkbBKBgFIwoAACEvQmxNvRKuAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-8860-5105","institution":"ShanghaiTech University","correspondingAuthor":true,"prefix":"","firstName":"Xufeng","middleName":"","lastName":"Kou","suffix":""},{"id":414154878,"identity":"75372045-0fd2-4f1b-863f-43e76755b670","order_by":1,"name":"Puyang Huang","email":"","orcid":"https://orcid.org/0000-0003-3013-9479","institution":"University of California, Los Angeles","correspondingAuthor":false,"prefix":"","firstName":"Puyang","middleName":"","lastName":"Huang","suffix":""},{"id":414154879,"identity":"49e0864d-79c3-4f3b-b720-626f7692be6d","order_by":2,"name":"Yu Gu","email":"","orcid":"","institution":"ShanghaiTech University","correspondingAuthor":false,"prefix":"","firstName":"Yu","middleName":"","lastName":"Gu","suffix":""},{"id":414154880,"identity":"302bd867-8014-4dcb-bcbb-46f13b298928","order_by":3,"name":"Chenyi Fu","email":"","orcid":"https://orcid.org/0009-0009-9495-0923","institution":"Beihang University","correspondingAuthor":false,"prefix":"","firstName":"Chenyi","middleName":"","lastName":"Fu","suffix":""},{"id":414154881,"identity":"ff7f23c0-60e6-41d8-a318-a4dce4a9bea6","order_by":4,"name":"Tianhao Chen","email":"","orcid":"","institution":"Beihang University","correspondingAuthor":false,"prefix":"","firstName":"Tianhao","middleName":"","lastName":"Chen","suffix":""},{"id":414154882,"identity":"5fd5da05-0a91-4c0a-acfc-1375fea430d3","order_by":5,"name":"Shan Yao","email":"","orcid":"","institution":"ShanghaiTech University","correspondingAuthor":false,"prefix":"","firstName":"Shan","middleName":"","lastName":"Yao","suffix":""},{"id":414154883,"identity":"896f0c49-b756-4b27-9117-ad23db77efad","order_by":6,"name":"Zhenghang Zhi","email":"","orcid":"","institution":"ShanghaiTech University","correspondingAuthor":false,"prefix":"","firstName":"Zhenghang","middleName":"","lastName":"Zhi","suffix":""},{"id":414154884,"identity":"d6de9e28-0a70-4570-95fd-2411e3eef210","order_by":7,"name":"Jiaqi Lu","email":"","orcid":"","institution":"Beihang University","correspondingAuthor":false,"prefix":"","firstName":"Jiaqi","middleName":"","lastName":"Lu","suffix":""},{"id":414154885,"identity":"64ba4924-9158-401c-a0fe-a1a3e042b91a","order_by":8,"name":"Yongqi Hu","email":"","orcid":"","institution":"ShanghaiTech University","correspondingAuthor":false,"prefix":"","firstName":"Yongqi","middleName":"","lastName":"Hu","suffix":""},{"id":414154886,"identity":"75f272ca-b347-4dbd-b6f6-76f2c327890f","order_by":9,"name":"Hongchao Zhang","email":"","orcid":"","institution":"Beihang University","correspondingAuthor":false,"prefix":"","firstName":"Hongchao","middleName":"","lastName":"Zhang","suffix":""},{"id":414154887,"identity":"db7e2cad-2ce6-4271-b9fb-c2d7c3f87d07","order_by":10,"name":"Shiyang Lu","email":"","orcid":"","institution":"Beihang University","correspondingAuthor":false,"prefix":"","firstName":"Shiyang","middleName":"","lastName":"Lu","suffix":""},{"id":414154888,"identity":"c7ddf7a4-9866-4f4b-a602-a2d976850477","order_by":11,"name":"Yumeng Yang","email":"","orcid":"https://orcid.org/0000-0002-8071-4470","institution":"ShanghaiTech University","correspondingAuthor":false,"prefix":"","firstName":"Yumeng","middleName":"","lastName":"Yang","suffix":""},{"id":414154889,"identity":"38e41797-cefe-4da7-84c3-8231e24ae022","order_by":12,"name":"Tianxiao Nie","email":"","orcid":"https://orcid.org/0000-0001-9067-9931","institution":"Beihang University","correspondingAuthor":false,"prefix":"","firstName":"Tianxiao","middleName":"","lastName":"Nie","suffix":""},{"id":414154890,"identity":"7fab8f3e-ccef-46e4-a808-43cb75e685ea","order_by":13,"name":"Shouzhong Peng","email":"","orcid":"","institution":"Beihang University","correspondingAuthor":false,"prefix":"","firstName":"Shouzhong","middleName":"","lastName":"Peng","suffix":""},{"id":414154891,"identity":"a1fbd140-3f62-4004-b063-8f38af53366a","order_by":14,"name":"Weisheng Zhao","email":"","orcid":"","institution":"Beihang University","correspondingAuthor":false,"prefix":"","firstName":"Weisheng","middleName":"","lastName":"Zhao","suffix":""},{"id":414154892,"identity":"86eb1069-b1f9-4193-b784-de193c8b56d2","order_by":15,"name":"Kang Wang","email":"","orcid":"https://orcid.org/0000-0002-9363-1279","institution":"University of California, Los Angeles","correspondingAuthor":false,"prefix":"","firstName":"Kang","middleName":"","lastName":"Wang","suffix":""}],"badges":[],"createdAt":"2025-01-26 10:10:20","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5905772/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5905772/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":79327036,"identity":"31061556-a4b5-4660-84fa-a43c20d3ecbc","added_by":"auto","created_at":"2025-03-27 05:41:30","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1768112,"visible":true,"origin":"","legend":"","description":"","filename":"SOTMRAMenablednoisetolerantandresourcesavingPBNN.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5905772/v1_covered_60b129b9-c1e1-43da-b06e-c8a1889ef9cc.pdf"},{"id":79326204,"identity":"2a865343-e156-4129-97f4-fbf744d1670b","added_by":"auto","created_at":"2025-03-27 05:33:27","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":860490,"visible":true,"origin":"","legend":"\u003cp\u003eSupplementary Information\u003c/p\u003e","description":"","filename":"SupplementaryInformation.docx","url":"https://assets-eu.researchsquare.com/files/rs-5905772/v1/d49a2a4b109614265ee4b11f.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"SOT-MRAM-enabled noise-tolerant and resource-saving probabilistic binary neural network","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-5905772/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5905772/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The deployment of large-scale full-precision neural networks is hindered by the escalating parameter sizes and inherent susceptibility to noise. Here, we present a non-volatile memory (NVM)-based probabilistic computing architecture, which harnesses both the stochastic nature and in-memory computing capabilities of NVM to enhance computation efficiency and robustness. Utilizing in-plane magnetized spin-orbit torque magneto-resistive random-access memory (i-SOT MRAM) cells, we demonstrate ultra-fast (400 ps), field-free write operations and voltage-controllable probabilistic states with a low variation of 3.8%. These features enable the implementation of an on-chip SOT probabilistic binary neural network (SOT-PBNN) hardware, achieving near-baseline accuracy (88.2%) on CIFAR-10 classification tasks. Moreover, the stochastic training process endows the SOT-PBNN to attain a 7-fold improvement in classification accuracy over full-precision networks under a 25% write/read noise, while reducing the parameter size by 1–2 orders of magnitude. Our work establishes a lightweight framework suitable for realizing artificial intelligence platforms on resource-constrained edge devices.","manuscriptTitle":"SOT-MRAM-enabled noise-tolerant and resource-saving probabilistic binary neural network","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-03-27 05:33:22","doi":"10.21203/rs.3.rs-5905772/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"nature-communications","isNatureJournal":true,"hasQc":false,"allowDirectSubmit":false,"externalIdentity":"NCOMMS","sideBox":"Learn more about [Nature Communications](http://www.nature.com/ncomms/)","snPcode":"","submissionUrl":"https://mts-ncomms.nature.com/","title":"Nature Communications","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"ejp","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":false}}],"origin":"","ownerIdentity":"9790a58e-3a8a-4909-abbf-c005d3d18103","owner":[],"postedDate":"March 27th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":44161719,"name":"Physical sciences/Materials science/Condensed-matter physics/Magnetic properties and materials"},{"id":44161720,"name":"Physical sciences/Materials science/Condensed-matter physics/Ferromagnetism"}],"tags":[],"updatedAt":"2025-03-27T05:33:22+00:00","versionOfRecord":[],"versionCreatedAt":"2025-03-27 05:33:22","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5905772","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5905772","identity":"rs-5905772","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}