Hardware-Adaptive and Superlinear-Capacity Memristor-based Associative Memory

Hardware-Adaptive and Superlinear-Capacity Memristor-based Associative Memory | 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 Hardware-Adaptive and Superlinear-Capacity Memristor-based Associative Memory Can Li, Chengping He, Mingrui Jiang, Keyi Shan, Szu-Hao Yang, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6560504/v1 This work is licensed under a CC BY 4.0 License Status: Published Journal Publication published 23 Feb, 2026 Read the published version in Nature Communications → Version 1 posted You are reading this latest preprint version Abstract Brain-inspired computing aims to mimic cognitive functions like associative memory, the ability to recall complete patterns from partial cues. Memristor technology offers promising hardware for such neuromorphic systems due to its potential for efficient in-memory analog computing. Hopfield Neural Networks (HNNs) are a classic model for associative memory, but implementations on conventional hardware suffer from efficiency bottlenecks, while prior memristor-based HNNs faced challenges with vulnerability to hardware defects due to offline training, limited storage capacity, and difficulty processing analog patterns. Here we introduce and experimentally demonstrate on integrated memristor hardware a new hardware-adaptive learning algorithm for associative memories that significantly improves defect tolerance and capacity, and naturally extends to scalable multilayer architectures capable of handling both binary and continuous patterns. Our approach achieves 3x effective capacity under 50% device faults compared to state-of-the-art methods. Furthermore, its extension to multilayer architectures enables superlinear capacity scaling (∝ N^1.49 for binary patterns) and effective recalling of continuous patterns (∝ N^1.74 scaling), as compared to linear capacity scaling for previous HNNs. It also provides flexibility to adjust capacity by tuning hidden neurons for the same-sized patterns. By leveraging the massive parallelism of the hardware enabled by synchronous updates, it reduces energy by 8.8× and latency by 99.7% for 64-dimensional patterns over asynchronous schemes, with greater improvements at scale. This promises the development of more reliable memristor-based associative memory systems and enables new applications research due to the significantly improved capacity, efficiency, and flexibility. Physical sciences/Engineering/Electrical and electronic engineering Physical sciences/Mathematics and computing/Computational science Full Text Additional Declarations There is NO Competing Interest. Supplementary Files AdaptiveassociatememorySI.pdf Supplementary Information Cite Share Download PDF Status: Published Journal Publication published 23 Feb, 2026 Read the published version in Nature Communications → 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-6560504","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":453810680,"identity":"feb4e9ea-0f0f-4659-9386-cd94662a5839","order_by":0,"name":"Can Li","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAlElEQVRIiWNgGAWjYHACxgMfGxIYGCTYSNBzcCbJWg7zkqSFv/3sgcO2O9ISG6TbEojTInEmL+Fw7pmcxAaZYweI02LAkGNwOLetIrFBIr2BSC38bwwOW5KmRQJoC2Mb0GESaUQ6TOLGG4ODvW1pxm0SaQnEaeHvzzF88LMtWbZfIs2AOC1wQEpEjoJRMApGwSggBAAIki2ZhBHEQgAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0003-3795-2008","institution":"The University of Hong Kong","correspondingAuthor":true,"prefix":"","firstName":"Can","middleName":"","lastName":"Li","suffix":""},{"id":453810681,"identity":"d446322b-0030-42d6-91d2-986f5b6a5f38","order_by":1,"name":"Chengping He","email":"","orcid":"","institution":"The University of Hong Kong","correspondingAuthor":false,"prefix":"","firstName":"Chengping","middleName":"","lastName":"He","suffix":""},{"id":453810682,"identity":"65588477-8696-4a8c-b3f8-b7ef09bd8405","order_by":2,"name":"Mingrui Jiang","email":"","orcid":"","institution":"The University of Hong Kong","correspondingAuthor":false,"prefix":"","firstName":"Mingrui","middleName":"","lastName":"Jiang","suffix":""},{"id":453810683,"identity":"23cbca88-8c23-4934-8f57-267fa6a53e4b","order_by":3,"name":"Keyi Shan","email":"","orcid":"","institution":"The University of Hong Kong","correspondingAuthor":false,"prefix":"","firstName":"Keyi","middleName":"","lastName":"Shan","suffix":""},{"id":453810684,"identity":"3ae3c2fb-853b-4c4f-b296-38141f9d441c","order_by":4,"name":"Szu-Hao Yang","email":"","orcid":"","institution":"The University of Hong Kong","correspondingAuthor":false,"prefix":"","firstName":"Szu-Hao","middleName":"","lastName":"Yang","suffix":""},{"id":453810685,"identity":"7911b042-3ebb-4c57-a21a-1e6e424ccf71","order_by":5,"name":"Zefan Li","email":"","orcid":"","institution":"The University of Hong Kong","correspondingAuthor":false,"prefix":"","firstName":"Zefan","middleName":"","lastName":"Li","suffix":""},{"id":453810686,"identity":"8b9862e6-9259-4ac2-a8bb-7cd76472a1c5","order_by":6,"name":"Shengbo Wang","email":"","orcid":"","institution":"The University of Hong Kong","correspondingAuthor":false,"prefix":"","firstName":"Shengbo","middleName":"","lastName":"Wang","suffix":""},{"id":453810687,"identity":"dd327153-1ab5-4bf8-8f59-2fd05518eaf6","order_by":7,"name":"Giacomo Pedretti","email":"","orcid":"https://orcid.org/0000-0002-4501-8672","institution":"Hewlett Packard Enterprise","correspondingAuthor":false,"prefix":"","firstName":"Giacomo","middleName":"","lastName":"Pedretti","suffix":""},{"id":453810688,"identity":"6b2d5ce3-7b7c-46fd-8ca8-f17d49e85121","order_by":8,"name":"Jim Ignowski","email":"","orcid":"","institution":"Hewlett Packard Labs","correspondingAuthor":false,"prefix":"","firstName":"Jim","middleName":"","lastName":"Ignowski","suffix":""}],"badges":[],"createdAt":"2025-04-30 03:06:00","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6560504/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6560504/v1","draftVersion":[],"editorialEvents":[{"content":"https://doi.org/10.1038/s41467-026-69958-0","type":"published","date":"2026-02-23T05:00:00+00:00"}],"editorialNote":"","failedWorkflow":false,"files":[{"id":105884450,"identity":"ff8c9cd3-7cfa-4712-80fb-d5c6362ea9f0","added_by":"auto","created_at":"2026-04-01 07:17:07","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2105361,"visible":true,"origin":"","legend":"Article File","description":"","filename":"AdaptiveassociatememoryMain.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6560504/v1_covered_57dd2c7e-e5d4-4dc8-9fb2-1366acd31f11.pdf"},{"id":82782494,"identity":"054017f0-91a2-4b0e-aa78-e16811589dc6","added_by":"auto","created_at":"2025-05-15 08:36:44","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":2611457,"visible":true,"origin":"","legend":"Supplementary Information","description":"","filename":"AdaptiveassociatememorySI.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6560504/v1/a567737d1e69678d09749355.pdf"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"Hardware-Adaptive and Superlinear-Capacity Memristor-based Associative Memory","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":true,"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-6560504/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6560504/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Brain-inspired computing aims to mimic cognitive functions like associative memory, the ability to recall complete patterns from partial cues. Memristor technology offers promising hardware for such neuromorphic systems due to its potential for efficient in-memory analog computing. Hopfield Neural Networks (HNNs) are a classic model for associative memory, but implementations on conventional hardware suffer from efficiency bottlenecks, while prior memristor-based HNNs faced challenges with vulnerability to hardware defects due to offline training, limited storage capacity, and difficulty processing analog patterns. Here we introduce and experimentally demonstrate on integrated memristor hardware a new hardware-adaptive learning algorithm for associative memories that significantly improves defect tolerance and capacity, and naturally extends to scalable multilayer architectures capable of handling both binary and continuous patterns. Our approach achieves 3x effective capacity under 50% device faults compared to state-of-the-art methods. Furthermore, its extension to multilayer architectures enables superlinear capacity scaling (∝ N^1.49 for binary patterns) and effective recalling of continuous patterns (∝ N^1.74 scaling), as compared to linear capacity scaling for previous HNNs. It also provides flexibility to adjust capacity by tuning hidden neurons for the same-sized patterns. By leveraging the massive parallelism of the hardware enabled by synchronous updates, it reduces energy by 8.8× and latency by 99.7% for 64-dimensional patterns over asynchronous schemes, with greater improvements at scale. This promises the development of more reliable memristor-based associative memory systems and enables new applications research due to the significantly improved capacity, efficiency, and flexibility.","manuscriptTitle":"Hardware-Adaptive and Superlinear-Capacity Memristor-based Associative Memory","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-05-15 08:36:32","doi":"10.21203/rs.3.rs-6560504/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":"0a45fac0-4538-444b-931d-fe64f67737bf","owner":[],"postedDate":"May 15th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"published-in-journal","subjectAreas":[{"id":48257837,"name":"Physical sciences/Engineering/Electrical and electronic engineering"},{"id":48257838,"name":"Physical sciences/Mathematics and computing/Computational science"}],"tags":[],"updatedAt":"2026-04-01T07:16:59+00:00","versionOfRecord":{"articleIdentity":"rs-6560504","link":"https://doi.org/10.1038/s41467-026-69958-0","journal":{"identity":"nature-communications","isVorOnly":false,"title":"Nature Communications"},"publishedOn":"2026-02-23 05:00:00","publishedOnDateReadable":"February 23rd, 2026"},"versionCreatedAt":"2025-05-15 08:36:32","video":"","vorDoi":"10.1038/s41467-026-69958-0","vorDoiUrl":"https://doi.org/10.1038/s41467-026-69958-0","workflowStages":[]},"version":"v1","identity":"rs-6560504","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6560504","identity":"rs-6560504","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}