Federated Reinforcement Learning Framework for Privacy Preserving Few Shot Learning

Federated Reinforcement Learning Framework for Privacy Preserving Few Shot Learning | 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 Federated Reinforcement Learning Framework for Privacy Preserving Few Shot Learning Minsoo Kang, Jihye Park, Donghyun Choi, Seoyeon Kim This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7616375/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 This study introduces a federated reinforcement learning framework for few-shot learning (FRL-FSL), aiming to address the dual challenges of data scarcity and privacy preservation in distributed environments. The proposed framework integrates policy gradient optimization with secure aggregation and introduces validator nodes to ensure the authenticity of both data and model updates. Experiments were conducted on the Omniglot and FC100 datasets under 1-shot and 5-shot conditions, with comparisons against FedAvg, FedFSL and traditional supervised baselines. Results demonstrate that FRL-FSL achieved an average accuracy of 87.3% on Omniglot (5-shot), improving by 25.9% over FedAvg and 13.8% over FedFSL, while maintaining 72.6% accuracy in 1-shot tasks. On the FC100 dataset, FRL-FSL reached 59.8% accuracy in 5-shot learning, outperforming FedAvg by 18.6% and FedFSL by 7.1%, and achieved 46.3% in 1-shot learning. The framework also reduced the privacy risk index by 37% relative to FedAvg, with convergence accelerated by nearly 30% compared to baselines. These findings confirm that FRL-FSL achieves a practical balance between accuracy, convergence, and privacy, offering a promising solution for real-world, privacy-sensitive applications. Artificial Intelligence and Machine Learning Theoretical Computer Science Federated learning Reinforcement learning Few-shot learning Privacy preservation Distributed optimization Omniglot FC100 Full Text Additional Declarations The authors declare no competing interests. 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-7616375","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":515026121,"identity":"34238483-2847-4212-a778-183dceb552d6","order_by":0,"name":"Minsoo Kang","email":"data:image/png;base64,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","orcid":"","institution":"Department of Computer Science and Engineering, Seoul National University, South Korea","correspondingAuthor":true,"prefix":"","firstName":"Minsoo","middleName":"","lastName":"Kang","suffix":""},{"id":515026122,"identity":"a3fc9495-4953-444c-9747-7e8993a81756","order_by":1,"name":"Jihye Park","email":"","orcid":"","institution":"School of Electrical and Computer Engineering, Korea Advanced Institute of Science and Technology (KAIST), South Korea","correspondingAuthor":false,"prefix":"","firstName":"Jihye","middleName":"","lastName":"Park","suffix":""},{"id":515026123,"identity":"04565527-322d-455c-a46a-cbc3790efdc7","order_by":2,"name":"Donghyun Choi","email":"","orcid":"","institution":"Department of Computer Science and Engineering, Seoul National University, South Korea","correspondingAuthor":false,"prefix":"","firstName":"Donghyun","middleName":"","lastName":"Choi","suffix":""},{"id":515026124,"identity":"a6caf379-d3bf-4aac-9a8b-b6994ec8b6a4","order_by":3,"name":"Seoyeon Kim","email":"","orcid":"","institution":"Department of Computer Science and Engineering, Seoul National University, South Korea","correspondingAuthor":false,"prefix":"","firstName":"Seoyeon","middleName":"","lastName":"Kim","suffix":""}],"badges":[],"createdAt":"2025-09-15 05:15:46","currentVersionCode":1,"declarations":{"humanSubjects":false,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":false,"humanSubjectConsent":false,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-7616375/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7616375/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":91398574,"identity":"8963f76a-5269-460d-9161-d9415fc9b53c","added_by":"auto","created_at":"2025-09-16 06:34:26","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":538797,"visible":true,"origin":"","legend":"","description":"","filename":"09151FederatedReinforcementLearningFrameworkforPrivacyPreservingFewShotLearning.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7616375/v1_covered_194d7f3b-2f6f-4cbe-90ef-e5b0821f831f.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eFederated Reinforcement Learning Framework for Privacy Preserving Few Shot Learning\u003c/strong\u003e\u003c/p\u003e","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"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":"Federated learning, Reinforcement learning, Few-shot learning, Privacy preservation, Distributed optimization, Omniglot, FC100","lastPublishedDoi":"10.21203/rs.3.rs-7616375/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7616375/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThis study introduces a federated reinforcement learning framework for few-shot learning (FRL-FSL), aiming to address the dual challenges of data scarcity and privacy preservation in distributed environments. The proposed framework integrates policy gradient optimization with secure aggregation and introduces validator nodes to ensure the authenticity of both data and model updates. Experiments were conducted on the Omniglot and FC100 datasets under 1-shot and 5-shot conditions, with comparisons against FedAvg, FedFSL and traditional supervised baselines. Results demonstrate that FRL-FSL achieved an average accuracy of 87.3% on Omniglot (5-shot), improving by 25.9% over FedAvg and 13.8% over FedFSL, while maintaining 72.6% accuracy in 1-shot tasks. On the FC100 dataset, FRL-FSL reached 59.8% accuracy in 5-shot learning, outperforming FedAvg by 18.6% and FedFSL by 7.1%, and achieved 46.3% in 1-shot learning. The framework also reduced the privacy risk index by 37% relative to FedAvg, with convergence accelerated by nearly 30% compared to baselines. These findings confirm that FRL-FSL achieves a practical balance between accuracy, convergence, and privacy, offering a promising solution for real-world, privacy-sensitive applications.\u003c/p\u003e","manuscriptTitle":"Federated Reinforcement Learning Framework for Privacy Preserving Few Shot Learning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-16 06:10:21","doi":"10.21203/rs.3.rs-7616375/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":"26b85db2-d166-42f3-ad48-1fb1ed14c7ec","owner":[],"postedDate":"September 16th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":54702017,"name":"Artificial Intelligence and Machine Learning"},{"id":54702018,"name":"Theoretical Computer Science"}],"tags":[],"updatedAt":"2025-09-16T06:10:21+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-16 06:10:21","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7616375","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7616375","identity":"rs-7616375","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}