Lightweight Edge Learning via Dataset Pruning

Lightweight Edge Learning via Dataset Pruning | 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 Lightweight Edge Learning via Dataset Pruning Laha Ale, Hu Luo, Mingsheng Cao, Shichao Li, Huanlai Xing, Haifeng Sun This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8637469/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 Edge learning facilitates ubiquitous intelligence by enabling model training and adaptation directly on data-generating devices, thereby mitigating privacy risks and communication latency. However, the high computational and energy overhead of on-device training hinders its deployment on battery-powered mobile systems with strict thermal and memory budgets. While prior research has extensively optimized model architectures for efficient inference, the training phase remains bottlenecked by the processing of massive, often redundant, local datasets. In this work, we propose a data-centric optimization framework that leverages dataset pruning to achieve resource-efficient edge learning. Unlike standard methods that process all available data, our approach constructs compact, highly informative training subsets via a lightweight, on-device importance evaluation. Specifically, we utilize average loss statistics derived from a truncated warm-up phase to rank sample importance, deterministically retaining only the most critical data points under a dynamic pruning ratio. This mechanism is model-agnostic and operates locally without inter-device communication. Extensive experiments on standard image classification benchmarks demonstrate that our framework achieves a near-linear reduction in training latency and energy consumption proportional to the pruning ratio, with negligible degradation in model accuracy. These results validate dataset pruning as a vital, complementary paradigm for enhancing the sustainability and scalability of learning on resource-constrained mobile edge devices. Physical sciences/Energy science and technology Physical sciences/Engineering Physical sciences/Mathematics and computing Mobile Edge Computing Data Pruning Edge Learning Full Text Additional Declarations No competing interests reported. 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-8637469","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":595740815,"identity":"79455d03-7cb2-47f3-bcd7-42a73d078173","order_by":0,"name":"Laha Ale","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA0UlEQVRIiWNgGAWjYBACPgYeEGXBwM9MrBY2iBYJBslmkrUYHCBaC/vZgx8YaiTkjI8zH3zwgcEmX96BkBaevGQJhmMSxmaH2ZINZzCkWW4kZB0bQ46BBAObROK2wzxm0jwMhw0MGwhp4X9j/IPhn0T95mb+77//EKVFIsdMgrFNIsGAmYcNGM6HDeQJ6ABqeWNmkdgnYTjjMJuxZI9BmoEBIS38/DnGNz58s5Hn7z/88MOPChsDeUIOA4MEOAtoBfERBAfE2TIKRsEoGAUjCQAA8QIzasyiEFgAAAAASUVORK5CYII=","orcid":"","institution":"Southwest Jiaotong University","correspondingAuthor":true,"prefix":"","firstName":"Laha","middleName":"","lastName":"Ale","suffix":""},{"id":595740816,"identity":"063651ad-35ce-4580-a849-847d734f5c14","order_by":1,"name":"Hu Luo","email":"","orcid":"","institution":"Southwest Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Hu","middleName":"","lastName":"Luo","suffix":""},{"id":595740817,"identity":"75183916-281f-45a1-9335-49c8d55d3a34","order_by":2,"name":"Mingsheng Cao","email":"","orcid":"","institution":"University of Electronic Science and Technology of China","correspondingAuthor":false,"prefix":"","firstName":"Mingsheng","middleName":"","lastName":"Cao","suffix":""},{"id":595740818,"identity":"32cc6f68-18e7-45e7-9171-d5220f3c8dc7","order_by":3,"name":"Shichao Li","email":"","orcid":"","institution":"Guilin University of Electronic Technology","correspondingAuthor":false,"prefix":"","firstName":"Shichao","middleName":"","lastName":"Li","suffix":""},{"id":595740819,"identity":"84ad5df9-2a5d-4797-ab3e-eb37a8929951","order_by":4,"name":"Huanlai Xing","email":"","orcid":"","institution":"Southwest Jiaotong University","correspondingAuthor":false,"prefix":"","firstName":"Huanlai","middleName":"","lastName":"Xing","suffix":""},{"id":595740820,"identity":"386c8ebb-8f69-4571-a9fe-2487ed6887d6","order_by":5,"name":"Haifeng Sun","email":"","orcid":"","institution":"Southwest University of Science and Technology","correspondingAuthor":false,"prefix":"","firstName":"Haifeng","middleName":"","lastName":"Sun","suffix":""}],"badges":[],"createdAt":"2026-01-19 09:23:16","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8637469/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8637469/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":103506639,"identity":"67994bfd-bcb6-4cd7-89d7-2fc3b4938174","added_by":"auto","created_at":"2026-02-26 13:38:22","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1513583,"visible":true,"origin":"","legend":"","description":"","filename":"LightweightEdgeLearningviaDatasetPruning.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8637469/v1_covered_adcb062e-6fde-487c-a688-82fac2243b49.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Lightweight Edge Learning via Dataset Pruning","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"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":"Mobile Edge Computing, Data Pruning, Edge Learning","lastPublishedDoi":"10.21203/rs.3.rs-8637469/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8637469/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":" \nEdge learning facilitates ubiquitous intelligence by enabling model training and adaptation directly on data-generating devices, thereby mitigating privacy risks and communication latency. However, the high computational and energy overhead of on-device training hinders its deployment on battery-powered mobile systems with strict thermal and memory budgets. While prior research has extensively optimized model architectures for efficient inference, the training phase remains bottlenecked by the processing of massive, often redundant, local datasets. In this work, we propose a data-centric optimization framework that leverages dataset pruning to achieve resource-efficient edge learning. Unlike standard methods that process all available data, our approach constructs compact, highly informative training subsets via a lightweight, on-device importance evaluation. Specifically, we utilize average loss statistics derived from a truncated warm-up phase to rank sample importance, deterministically retaining only the most critical data points under a dynamic pruning ratio. This mechanism is model-agnostic and operates locally without inter-device communication. Extensive experiments on standard image classification benchmarks demonstrate that our framework achieves a near-linear reduction in training latency and energy consumption proportional to the pruning ratio, with negligible degradation in model accuracy. These results validate dataset pruning as a vital, complementary paradigm for enhancing the sustainability and scalability of learning on resource-constrained mobile edge devices.","manuscriptTitle":"Lightweight Edge Learning via Dataset Pruning","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-25 05:43:12","doi":"10.21203/rs.3.rs-8637469/v1","editorialEvents":[],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"communications-ai-and-computing","isNatureJournal":true,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"","sideBox":"Learn more about [Communications AI \u0026 Computing](https://www.nature.com/commsaicomp/)","snPcode":"44488","submissionUrl":"https://submission.nature.com/new-submission/44488/3","title":"Communications AI \u0026 Computing","twitterHandle":"","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"Nature Communications","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"03d145e3-a7ab-4bc3-9eb2-b99ac59bc7bf","owner":[],"postedDate":"February 25th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[{"id":63390146,"name":"Physical sciences/Energy science and technology"},{"id":63390147,"name":"Physical sciences/Engineering"},{"id":63390148,"name":"Physical sciences/Mathematics and computing"}],"tags":[],"updatedAt":"2026-02-25T05:43:12+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-25 05:43:12","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8637469","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8637469","identity":"rs-8637469","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}