Relation Graph Alignment for Learning with Noisy Labels

Relation Graph Alignment for Learning with Noisy Labels | 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 Research Article Relation Graph Alignment for Learning with Noisy Labels Zhuang Qi, Junlin Zhang, Sijin Zhou, Xiaming Chen, Xin Qi This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6233232/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Noisy label learning is crucial for improving model performance on datasets with unreliable labels, which enables models to better tolerate inaccurate data. Existing methods either focus on optimizing the loss function to mitigate the interference from noise, or design procedures to detect potential noise and correct errors. However, their effectiveness is often compromised in representation learning due to the dilemma where models overfit to noisy labels. To address the aforementioned issue, this paper proposes a relation graph alignment framework that models inter-sample relationships via self-supervised learning and employs knowledge distillation to enhance understanding of latent associations, which mitigate the impact of noisy labels. Specifically, the proposed method, termed RMDNet, includes two main modules, where the relation modeling (RM) module implements the contrastive learning technique to learn representations of all data, an unsupervised approach that effectively eliminates the interference of noisy tags on feature extraction. The relation-guided representation learning (RGRL) module utilizes inter-sample relation learned from the RM module to calibrate the representation distribution for noisy samples, which is capable of improving the generalization of the model in the inference phase. Notably, the proposed RMDNet is a plug-and-play framework that can integrate multiple methods to its advantage. The RMDNet can achieve a performance improvement of 1% to 8% when combined with existing methods and learn discriminative representations for noisy data, which results in superior performance than the existing methods. Noisy label learning Relation modeling Knowledge distillation Full Text Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted 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-6233232","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":466816122,"identity":"2bc4bee9-6738-4009-8533-b03de2dce314","order_by":0,"name":"Zhuang Qi","email":"","orcid":"","institution":"Shandong University","correspondingAuthor":false,"prefix":"","firstName":"Zhuang","middleName":"","lastName":"Qi","suffix":""},{"id":466816123,"identity":"7b9b12a1-9ce9-4a17-86ec-3023a3e0a8a3","order_by":1,"name":"Junlin Zhang","email":"","orcid":"","institution":"Shantou University","correspondingAuthor":false,"prefix":"","firstName":"Junlin","middleName":"","lastName":"Zhang","suffix":""},{"id":466816124,"identity":"7c53e5ee-bd5e-457d-bbb7-080f76856da2","order_by":2,"name":"Sijin Zhou","email":"","orcid":"","institution":"Monash University","correspondingAuthor":false,"prefix":"","firstName":"Sijin","middleName":"","lastName":"Zhou","suffix":""},{"id":466816125,"identity":"0186b522-d30e-4496-9e26-96f147872026","order_by":3,"name":"Xiaming Chen","email":"","orcid":"","institution":"Shantou University","correspondingAuthor":false,"prefix":"","firstName":"Xiaming","middleName":"","lastName":"Chen","suffix":""},{"id":466816126,"identity":"9a7715c1-c5d6-4a35-b683-85e13fb2ea41","order_by":4,"name":"Xin Qi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAxElEQVRIiWNgGAWjYDACZgaGAwwVUA4P8VrOQFUTpwUEGNtI0cJ3nMfwcOG8w/b2/AcYH7xtY5A3J6RF8jBbwuGZ2w4n9kgkMBvObWMw3NlAQIvBYeYDh3m3HU7gkWBgk+ZtY0gwOEBQC2PDYd45h+15+A+w/yZSC8iWhsOMPQwJbMxEaQH7hedYemLPjcRmyTnnJAw3ENLCd/6M8WeeGmt79v7DBz+8KbORJ2gLA0IBYwOQkCCkHkXLKBgFo2AUjAIcAABfxTz4ZaSvOAAAAABJRU5ErkJggg==","orcid":"","institution":"Suzhou University of Science and Technology","correspondingAuthor":true,"prefix":"","firstName":"Xin","middleName":"","lastName":"Qi","suffix":""}],"badges":[],"createdAt":"2025-03-15 13:53:10","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6233232/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6233232/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":85463901,"identity":"da9cf61b-9e46-4c6e-a598-5515dfa042b1","added_by":"auto","created_at":"2025-06-26 08:02:11","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":781722,"visible":true,"origin":"","legend":"","description":"","filename":"MMSJ.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6233232/v1_covered_22cc2d35-0f74-48bb-80f9-743b56cbf4ca.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Relation Graph Alignment for Learning with Noisy Labels","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"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":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Noisy label learning, Relation modeling, Knowledge distillation","lastPublishedDoi":"10.21203/rs.3.rs-6233232/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6233232/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Noisy label learning is crucial for improving model performance on datasets with unreliable labels, which enables models to better tolerate inaccurate data. Existing methods either focus on optimizing the loss function to mitigate the interference from noise, or design procedures to detect potential noise and correct errors. However, their effectiveness is often compromised in representation learning due to the dilemma where models overfit to noisy labels. To address the aforementioned issue, this paper proposes a relation graph alignment framework that models inter-sample relationships via self-supervised learning and employs knowledge distillation to enhance understanding of latent associations, which mitigate the impact of noisy labels. Specifically, the proposed method, termed RMDNet, includes two main modules, where the relation modeling (RM) module implements the contrastive learning technique to learn representations of all data, an unsupervised approach that effectively eliminates the interference of noisy tags on feature extraction. The relation-guided representation learning (RGRL) module utilizes inter-sample relation learned from the RM module to calibrate the representation distribution for noisy samples, which is capable of improving the generalization of the model in the inference phase. Notably, the proposed RMDNet is a plug-and-play framework that can integrate multiple methods to its advantage. The RMDNet can achieve a performance improvement of 1% to 8% when combined with existing methods and learn discriminative representations for noisy data, which results in superior performance than the existing methods.","manuscriptTitle":"Relation Graph Alignment for Learning with Noisy Labels","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-06 08:23:40","doi":"10.21203/rs.3.rs-6233232/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"1c1ee848-1a56-4701-a551-c068c5fcfe21","owner":[],"postedDate":"June 6th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-06-26T07:54:04+00:00","versionOfRecord":[],"versionCreatedAt":"2025-06-06 08:23:40","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6233232","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6233232","identity":"rs-6233232","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}