{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,5]],"date-time":"2026-02-05T06:47:50Z","timestamp":1770274070499,"version":"3.49.0"},"reference-count":32,"publisher":"IOP Publishing","issue":"1","license":[{"start":{"date-parts":[[2025,1,31]],"date-time":"2025-01-31T00:00:00Z","timestamp":1738281600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2025,1,31]],"date-time":"2025-01-31T00:00:00Z","timestamp":1738281600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/iopscience.iop.org\/info\/page\/text-and-data-mining"}],"content-domain":{"domain":["iopscience.iop.org"],"crossmark-restriction":false},"short-container-title":["Mach. Learn.: Sci. Technol."],"published-print":{"date-parts":[[2025,3,31]]},"abstract":"<jats:title>Abstract<\/jats:title>\n               <jats:p>This paper introduces an innovative method for fine-tuning a large multi-label model for abnormality detection, utilizing a smaller trainer and advanced knowledge distillation techniques. It studies the effects of fine-tuning on various abnormalities, noting different improvements based on the Original Model\u2019s performance in specific tasks. The experimental setup, optimized for on-device inference and fine-tuning with limited computational resources, demonstrates moderate yet promising enhancements in model performance post-fine-tuning. Key insights from the study include the significance of aligning the <jats:italic>\u00b5<\/jats:italic>-Trainer\u2019s behavior with the Original Model and the influence of hyper-parameters like batch size on fine-tuning outcomes. The research acknowledges limitations such as the limited exploration of loss functions in multi-label models and constraints in architectural design, suggesting potential avenues for future investigation. While the proposed <jats:italic>Naive Continual Fine-tuning Process<\/jats:italic> is in its early stages, we highlight this paper\u2019s potential model personalization on long-term data. Moreover, weight transfer in our system is exclusively for fine-tuning; hence, it improves user privacy protection by failing data reconstruction attempts from weights, like an issue with Federated learning models. Our on-device fine-tuning prevents the transferring of data or gradients from the edge of the network to their server. Despite modest performance improvements after fine-tuning, these working layers represent a small fraction (0.7%) of the total weights in the Original Model and 1.6% in the <jats:italic>\u00b5<\/jats:italic>-Trainer. This study establishes a foundational framework for advancing personalized model adaptation, on-device inference and fine-tuning while emphasizing the importance of safeguarding data privacy in model development.<\/jats:p>","DOI":"10.1088\/2632-2153\/adaca3","type":"journal-article","created":{"date-parts":[[2025,1,21]],"date-time":"2025-01-21T22:55:46Z","timestamp":1737500146000},"page":"015025","update-policy":"https:\/\/doi.org\/10.1088\/crossmark-policy","source":"Crossref","is-referenced-by-count":3,"title":["Advancing privacy-aware machine learning on sensitive data via edge-based continual \u00b5-training for personalized large models"],"prefix":"10.1088","volume":"6","author":[{"ORCID":"https:\/\/orcid.org\/0009-0004-2796-6734","authenticated-orcid":true,"given":"Zhaojing","family":"Huang","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0008-4794-9586","authenticated-orcid":false,"given":"Leping","family":"Yu","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-8458-9486","authenticated-orcid":true,"given":"Luis","family":"Fernando Herbozo Contreras","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0007-5372-8010","authenticated-orcid":false,"given":"Kamran","family":"Eshraghian","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4350-8026","authenticated-orcid":false,"given":"Nhan","family":"Duy Truong","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2384-0710","authenticated-orcid":false,"given":"Armin","family":"Nikpour","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2753-5553","authenticated-orcid":true,"given":"Omid","family":"Kavehei","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"266","published-online":{"date-parts":[[2025,1,31]]},"reference":[{"key":"mlstadaca3bib1","first-page":"p 32","article-title":"Deep leakage from gradients","author":"Zhu","year":"2019"},{"key":"mlstadaca3bib2","first-page":"pp 241","article-title":"Reconstructing individual data points in federated learning hardened with differential privacy and secure aggregation","author":"Boenisch","year":"2023"},{"key":"mlstadaca3bib3","doi-asserted-by":"publisher","first-page":"413","DOI":"10.1002\/clc.22667","article-title":"Silent atrial fibrillation: epidemiology, diagnosis and clinical impact","volume":"40","author":"Dilaveris","year":"2017","journal-title":"Clin. Cardiol."},{"key":"mlstadaca3bib4","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1007\/s13239-024-00716-3","article-title":"S4D-ECG: a shallow state-of-the-art model for cardiac abnormality classification","volume":"15","author":"Huang","year":"2024","journal-title":"Cardiovasc. Eng. Technol."},{"key":"mlstadaca3bib5","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1007\/s12265-024-10504-y","article-title":"Efficient edge-AI models for robust ECG abnormality detection on resource-constrained hardware","volume":"17","author":"Huang","year":"2024","journal-title":"J. Cardiovasc. Transl. Res."},{"key":"mlstadaca3bib6","doi-asserted-by":"publisher","DOI":"10.1063\/5.0191571","article-title":"On-device edge-learning for cardiac abnormality detection using a bio-inspired and spiking shallow network","volume":"2","author":"Huang","year":"2024","journal-title":"APL Mach. Learn."},{"key":"mlstadaca3bib7","doi-asserted-by":"publisher","DOI":"10.1063\/5.0191574","article-title":"Cardiac abnormality detection with a tiny diagonal state space model based on sequential liquid neural processing unit","volume":"2","author":"Huang","year":"2024","journal-title":"APL Mach. Learn."},{"key":"mlstadaca3bib8","article-title":"Security and privacy for artificial intelligence: opportunities and challenges","author":"Oseni","year":"2021"},{"key":"mlstadaca3bib9","first-page":"pp 1068","article-title":"Distilled deep learning based classification of abnormal heartbeat using ECG data through a low cost edge device","author":"Hartmann","year":"2019"},{"key":"mlstadaca3bib10","first-page":"pp 417","article-title":"A machine learning framework for edge computing to improve prediction accuracy in mobile health monitoring","volume":"vol 19","author":"Shree Ram","year":"2019"},{"key":"mlstadaca3bib11","doi-asserted-by":"publisher","first-page":"2533","DOI":"10.3390\/s20092533","article-title":"Edge machine learning for AI-enabled IoT devices: a review","volume":"20","author":"Merenda","year":"2020","journal-title":"Sensors"},{"key":"mlstadaca3bib12","doi-asserted-by":"publisher","DOI":"10.1016\/j.cie.2020.106854","article-title":"A review of applications in federated learning","volume":"149","author":"Li","year":"2020","journal-title":"Comput. Ind. Eng."},{"key":"mlstadaca3bib13","doi-asserted-by":"publisher","first-page":"778","DOI":"10.1109\/JBHI.2022.3181823","article-title":"Federated learning for privacy preservation in smart healthcare systems: a comprehensive survey","volume":"27","author":"Ali","year":"2022","journal-title":"IEEE J. Biomed. Health Inform."},{"key":"mlstadaca3bib14","doi-asserted-by":"publisher","first-page":"1760","DOI":"10.1038\/s41467-020-15432-4","article-title":"Automatic diagnosis of the 12-lead ECG using a deep neural network","volume":"11","author":"Ribeiro","year":"2020","journal-title":"Nat. Commun."},{"key":"mlstadaca3bib15","article-title":"Distilling the knowledge in a neural network","author":"Hinton","year":"2015"},{"key":"mlstadaca3bib16","article-title":"Offsite-tuning: Transfer learning without full model","author":"Xiao","year":"2023"},{"key":"mlstadaca3bib17","doi-asserted-by":"publisher","first-page":"1368","DOI":"10.1166\/jmihi.2018.2442","article-title":"An open access database for evaluating the algorithms of electrocardiogram rhythm and morphology abnormality detection","volume":"8","author":"Liu","year":"2018","journal-title":"J. Med. Imaging Health Inform."},{"key":"mlstadaca3bib18","doi-asserted-by":"publisher","first-page":"885","DOI":"10.1016\/j.jpeds.2013.03.054","article-title":"Long-term outcomes of pediatric sinus bradycardia","volume":"163","author":"Chiu","year":"2013","journal-title":"J. Ped."},{"key":"mlstadaca3bib19","doi-asserted-by":"publisher","first-page":"793","DOI":"10.1016\/j.amjcard.2018.05.015","article-title":"The disease burden of atrial fibrillation in China from a national cross-sectional survey","volume":"122","author":"Wang","year":"2018","journal-title":"Am. J. Cardiol."},{"key":"mlstadaca3bib20","doi-asserted-by":"publisher","first-page":"104","DOI":"10.1016\/j.eupc.2004.12.007","article-title":"Prevalence, characteristics and natural course of inappropriate sinus tachycardia","volume":"7","author":"Still","year":"2005","journal-title":"EP Eur."},{"key":"mlstadaca3bib21","doi-asserted-by":"publisher","first-page":"604","DOI":"10.1002\/clc.22454","article-title":"The prognostic significance of right bundle branch block: a meta-analysis of prospective cohort studies","volume":"38","author":"Xiong","year":"2015","journal-title":"Clin. Cardiol."},{"key":"mlstadaca3bib22","doi-asserted-by":"publisher","first-page":"181","DOI":"10.1016\/j.jacep.2016.02.012","article-title":"Outcomes related to first-degree atrioventricular block and therapeutic implications in patients with heart failure","volume":"2","author":"Nikolaidou","year":"2016","journal-title":"JACC: Clin. Electrophysiol."},{"key":"mlstadaca3bib23","doi-asserted-by":"publisher","DOI":"10.1111\/anec.12572","article-title":"Left bundle branch block: epidemiology, etiology, anatomic features, electrovectorcardiography and classification proposal","volume":"24","author":"P\u00e9rez-Riera","year":"2019","journal-title":"Ann. Noninvasive Electrocardiol."},{"key":"mlstadaca3bib24","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1080\/14796678.2024.2354082","article-title":"Generalization challenges in electrocardiogram deep learning: insights from dataset characteristics and attention mechanism","volume":"20","author":"Huang","year":"2024","journal-title":"Future Cardiol."},{"key":"mlstadaca3bib25","first-page":"pp 1176","article-title":"Federated learning for arrhythmia detection of non-iid ecg","author":"Zhang","year":"2020"},{"key":"mlstadaca3bib26","first-page":"pp 277","article-title":"Asynchronous federated learning-based ecg analysis for arrhythmia detection","author":"Sakib","year":"2021"},{"key":"mlstadaca3bib27","first-page":"pp 18","article-title":"Shake to leak: Fine-tuning diffusion models can amplify the generative privacy risk","author":"Zhangheng","year":"2024"},{"key":"mlstadaca3bib28","article-title":"Risks when sharing lora fine-tuned diffusion model weights","author":"Yao","year":"2024"},{"key":"mlstadaca3bib29","doi-asserted-by":"publisher","first-page":"94","DOI":"10.3390\/fi13040094","article-title":"Privacy preserving machine learning with homomorphic encryption and federated learning","volume":"13","author":"Fang","year":"2021","journal-title":"Future Internet"},{"key":"mlstadaca3bib30","first-page":"pp 1005","article-title":"Secure multiparty computation for privacy preserving data mining","author":"Lindell","year":"2005"},{"key":"mlstadaca3bib31","first-page":"pp 308","article-title":"Deep learning with differential privacy","author":"Abadi","year":"2016"},{"key":"mlstadaca3bib32","first-page":"pp 1322","article-title":"Model inversion attacks that exploit confidence information and basic countermeasures","author":"Fredrikson","year":"2015"}],"container-title":["Machine Learning: Science and Technology"],"original-title":[],"link":[{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adaca3","content-type":"text\/html","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adaca3\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adaca3","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adaca3\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adaca3\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adaca3\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"syndication"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adaca3\/pdf","content-type":"application\/pdf","content-version":"am","intended-application":"similarity-checking"},{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adaca3\/pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,1,31]],"date-time":"2025-01-31T13:36:53Z","timestamp":1738330613000},"score":1,"resource":{"primary":{"URL":"https:\/\/iopscience.iop.org\/article\/10.1088\/2632-2153\/adaca3"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,1,31]]},"references-count":32,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2025,1,31]]},"published-print":{"date-parts":[[2025,3,31]]}},"URL":"https:\/\/doi.org\/10.1088\/2632-2153\/adaca3","relation":{},"ISSN":["2632-2153"],"issn-type":[{"value":"2632-2153","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,1,31]]},"assertion":[{"value":"Advancing privacy-aware machine learning on sensitive data via edge-based continual \u00b5-training for personalized large models","name":"article_title","label":"Article Title"},{"value":"Machine Learning: Science and Technology","name":"journal_title","label":"Journal Title"},{"value":"paper","name":"article_type","label":"Article Type"},{"value":"\u00a9 2025 The Author(s). Published by IOP Publishing Ltd","name":"copyright_information","label":"Copyright Information"},{"value":"2024-05-18","name":"date_received","label":"Date Received","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2025-01-20","name":"date_accepted","label":"Date Accepted","group":{"name":"publication_dates","label":"Publication dates"}},{"value":"2025-01-31","name":"date_epub","label":"Online publication date","group":{"name":"publication_dates","label":"Publication dates"}}]}}