{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,9]],"date-time":"2026-07-09T06:41:45Z","timestamp":1783579305413,"version":"3.55.0"},"reference-count":54,"publisher":"Association for Computing Machinery (ACM)","issue":"4","license":[{"start":{"date-parts":[[2024,6,10]],"date-time":"2024-06-10T00:00:00Z","timestamp":1717977600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/www.acm.org\/publications\/policies\/copyright_policy#Background"}],"funder":[{"name":"NEPHELE","award":["101070487"],"award-info":[{"award-number":["101070487"]}]},{"name":"SMARTEDGE","award":["101092908"],"award-info":[{"award-number":["101092908"]}]},{"name":"European Union\u2019s Horizon Europe research and innovation"}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Embed. Comput. Syst."],"published-print":{"date-parts":[[2024,7,31]]},"abstract":"<jats:p>Recent advances in Tiny Machine Learning (TinyML) empower low-footprint embedded devices for real-time on-device Machine Learning (ML). While many acknowledge the potential benefits of TinyML, its practical implementation presents unique challenges. This study aims to bridge the gap between prototyping single TinyML models and developing reliable TinyML systems in production: (1)\u00a0Embedded devices operate in dynamically changing conditions. Existing TinyML solutions primarily focus on inference, with models trained offline on powerful machines and deployed as static objects. However, static models may underperform in the real world due to evolving input data distributions. We propose online learning to enable training on constrained devices, adapting local models toward the latest field conditions. (2)\u00a0Nevertheless, current on-device learning methods struggle with heterogeneous deployment conditions and the scarcity of labeled data when applied across numerous devices. We introduce federated meta-learning incorporating online learning to enhance model generalization, facilitating rapid learning. This approach ensures optimal performance among distributed devices by knowledge sharing. (3)\u00a0Moreover, TinyML\u2019s pivotal advantage is widespread adoption. Embedded devices and TinyML models prioritize extreme efficiency, leading to diverse characteristics ranging from memory and sensors to model architectures. Given their diversity and non-standardized representations, managing these resources becomes challenging as TinyML systems scale up. We present semantic management for the joint management of models and devices at scale. We demonstrate our methods through a basic regression example and then assess them in three real-world TinyML applications: handwritten character image classification, keyword audio classification, and smart building presence detection. The results confirm the effectiveness of our approaches from various perspectives, such as accuracy improvement, resource savings, and engineering effort reduction.<\/jats:p>","DOI":"10.1145\/3665278","type":"journal-article","created":{"date-parts":[[2024,5,16]],"date-time":"2024-05-16T11:25:23Z","timestamp":1715858723000},"page":"1-32","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":19,"title":["On-device Online Learning and Semantic Management of TinyML Systems"],"prefix":"10.1145","volume":"23","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-0241-6507","authenticated-orcid":false,"given":"Haoyu","family":"Ren","sequence":"first","affiliation":[{"name":"Siemens AG, Munich, Germany and Technical University of Munich, Garching bei Munich, Germany"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0583-4376","authenticated-orcid":false,"given":"Darko","family":"Anicic","sequence":"additional","affiliation":[{"name":"Siemens AG, Munich, Germany"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4515-6792","authenticated-orcid":false,"given":"Xue","family":"Li","sequence":"additional","affiliation":[{"name":"The University of Queensland, Brisbane, Australia"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-5465-198X","authenticated-orcid":false,"given":"Thomas","family":"Runkler","sequence":"additional","affiliation":[{"name":"Siemens AG, Munich, Germany and Technical University of Munich, Garching bei Munich, Germany"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"320","published-online":{"date-parts":[[2024,6,10]]},"reference":[{"key":"e_1_3_2_2_2","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.2107.01105"},{"key":"e_1_3_2_3_2","first-page":"1","volume-title":"Proceedings of the IEEE International Conference on Evolving and Adaptive Intelligent Systems (EAIS\u201922)","author":"Antonini Mattia","year":"2022","unstructured":"Mattia Antonini, Miguel Pincheira, Massimo Vecchio, and Fabio Antonelli. 2022. Tiny-MLOps: A framework for orchestrating ML applications at the far edge of IoT systems. In Proceedings of the IEEE International Conference on Evolving and Adaptive Intelligent Systems (EAIS\u201922). IEEE, 1\u20138."},{"key":"e_1_3_2_4_2","unstructured":"Antreas Antoniou Harrison Edwards and Amos Storkey. 2019. How to Train Your MAML. Retrieved from https:\/\/arxiv:1810.09502"},{"key":"e_1_3_2_5_2","doi-asserted-by":"publisher","unstructured":"Colby Banbury Vijay Janapa Reddi Peter Torelli Jeremy Holleman Nat Jeffries Csaba Kiraly Pietro Montino David Kanter Sebastian Ahmed Danilo Pau Urmish Thakker Antonio Torrini Peter Warden Jay Cordaro Giuseppe Di Guglielmo Javier Duarte Stephen Gibellini Videet Parekh Honson Tran Nhan Tran Niu Wenxu and Xu Xuesong. 2021. MLPerf Tiny Benchmark. Retrieved from https:\/\/arxiv.org\/abs\/2106.07597. DOI:10.48550\/ARXIV.2106.07597","DOI":"10.48550\/ARXIV.2106.07597"},{"key":"e_1_3_2_6_2","first-page":"11285","article-title":"Tinytl: Reduce memory, not parameters for efficient on-device learning","volume":"33","author":"Cai Han","year":"2020","unstructured":"Han Cai, Chuang Gan, Ligeng Zhu, and Song Han. 2020. Tinytl: Reduce memory, not parameters for efficient on-device learning. Adv. Neural Info. Process. Syst. 33 (2020), 11285\u201311297.","journal-title":"Adv. Neural Info. Process. Syst."},{"key":"e_1_3_2_7_2","first-page":"55","volume-title":"Proceedings of the International Semantic Web Conference","author":"Charpenay Victor","year":"2016","unstructured":"Victor Charpenay, Sebastian K\u00e4bisch, and Harald Kosch. 2016. Introducing thing descriptions and interactions: An ontology for the Web of Things. In Proceedings of the International Semantic Web Conference. Springer, 55\u201366."},{"key":"e_1_3_2_8_2","doi-asserted-by":"publisher","DOI":"10.1109\/icnc52316.2021.9607968"},{"key":"e_1_3_2_9_2","first-page":"800","article-title":"Tensorflow lite micro: Embedded machine learning for tinyML systems","volume":"3","author":"David Robert","year":"2021","unstructured":"Robert David, Jared Duke, Advait Jain, Vijay Janapa Reddi, Nat Jeffries, Jian Li, Nick Kreeger, Ian Nappier, Meghna Natraj, Tiezhen Wang et\u00a0al. 2021. Tensorflow lite micro: Embedded machine learning for tinyML systems. Proc. Mach. Learn. Syst. 3 (2021), 800\u2013811.","journal-title":"Proc. Mach. Learn. Syst."},{"key":"e_1_3_2_10_2","unstructured":"Giorgia Dellaferrera and Gabriel Kreiman. 2023. Error-driven Input Modulation: Solving the Credit Assignment Problem without a Backward Pass. Retrieved from https:\/\/arxiv:2201.11665"},{"key":"e_1_3_2_11_2","doi-asserted-by":"publisher","DOI":"10.1145\/3417313.3429378"},{"key":"e_1_3_2_12_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-48869-1_5"},{"key":"e_1_3_2_13_2","doi-asserted-by":"publisher","DOI":"10.1109\/VLSI-DAT52063.2021.9427352"},{"key":"e_1_3_2_14_2","doi-asserted-by":"publisher","unstructured":"Alireza Fallah Aryan Mokhtari and Asuman Ozdaglar. 2020. Personalized Federated Learning: A Meta-Learning Approach. Retrieved from https:\/\/arxiv.org\/abs\/2002.07948. DOI:10.48550\/ARXIV.2002.07948","DOI":"10.48550\/ARXIV.2002.07948"},{"key":"e_1_3_2_15_2","first-page":"1126","volume-title":"Proceedings of the 34th International Conference on Machine Learnin (ICML\u201917)","author":"Finn Chelsea","year":"2017","unstructured":"Chelsea Finn, Pieter Abbeel, and Sergey Levine. 2017. Model-agnostic meta-learning for fast adaptation of deep networks. In Proceedings of the 34th International Conference on Machine Learnin (ICML\u201917). JMLR.org, 1126\u20131135."},{"key":"e_1_3_2_16_2","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1902.08438"},{"key":"e_1_3_2_17_2","unstructured":"Jorge Gomez Saavan Patel Syed Shakib Sarwar Ziyun Li Raffaele Capoccia Zhao Wang Reid Pinkham Andrew Berkovich Tsung-Hsun Tsai Barbara De Salvo and Chiao Liu. 2022. Distributed On-Sensor Compute System for AR\/VR Devices: A Semi-Analytical Simulation Framework for Power Estimation. Retrieved from https:\/\/arxiv:2203.07474"},{"key":"e_1_3_2_18_2","doi-asserted-by":"publisher","DOI":"10.1145\/3462203.3475896"},{"key":"e_1_3_2_19_2","doi-asserted-by":"publisher","DOI":"10.23919\/ursi.2018.8406745"},{"key":"e_1_3_2_20_2","doi-asserted-by":"publisher","DOI":"10.1109\/icaiic54071.2022.9722636"},{"key":"e_1_3_2_21_2","unstructured":"Pouya Houshmand Jiacong Sun and Marian Verhelst. 2023. Benchmarking and Modeling of Analog and Digital SRAM in-memory Computing Architectures. Retrieved from https:\/\/arxiv:2305.18335"},{"key":"e_1_3_2_22_2","unstructured":"Shawn Hymel Colby Banbury Daniel Situnayake Alex Elium Carl Ward Mat Kelcey Mathijs Baaijens Mateusz Majchrzycki Jenny Plunkett David Tischler Alessandro Grande Louis Moreau Dmitry Maslov Artie Beavis Jan Jongboom and Vijay Janapa Reddi. 2023. Edge Impulse: An MLOps Platform for Tiny Machine Learning. Retrieved from https:\/\/arxiv:2212.03332"},{"key":"e_1_3_2_23_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.websem.2018.06.003"},{"key":"e_1_3_2_24_2","doi-asserted-by":"publisher","DOI":"10.1145\/3609121"},{"key":"e_1_3_2_25_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.cobeha.2019.04.007"},{"key":"e_1_3_2_26_2","first-page":"22941","article-title":"On-device training under 256kb memory","volume":"35","author":"Lin Ji","year":"2022","unstructured":"Ji Lin, Ligeng Zhu, Wei-Ming Chen, Wei-Chen Wang, Chuang Gan, and Song Han. 2022. On-device training under 256kb memory. Adv. Neural Info. Process. Syst. 35 (2022), 22941\u201322954.","journal-title":"Adv. Neural Info. Process. Syst."},{"key":"e_1_3_2_27_2","doi-asserted-by":"publisher","DOI":"10.1109\/icdcs47774.2020.00032"},{"key":"e_1_3_2_28_2","unstructured":"Yejia Liu Shijin Duan Xiaolin Xu and Shaolei Ren. 2023. MetaLDC: Meta Learning of Low-Dimensional Computing Classifiers for Fast On-Device Adaption. Retrieved from https:\/\/arxiv:2302.12347"},{"key":"e_1_3_2_29_2","unstructured":"Ilya Loshchilov and Frank Hutter. 2017. SGDR: Stochastic Gradient Descent with Warm Restarts. Retrieved from https:\/\/arxiv:1608.03983"},{"key":"e_1_3_2_30_2","doi-asserted-by":"publisher","DOI":"10.1109\/MeMeA57477.2023.10171940"},{"key":"e_1_3_2_31_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-68204-4_29"},{"key":"e_1_3_2_32_2","doi-asserted-by":"publisher","DOI":"10.1145\/3287560.3287596"},{"key":"e_1_3_2_33_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-65384-2_3"},{"key":"e_1_3_2_34_2","unstructured":"Alex Nichol Joshua Achiam and John Schulman. 2018. On First-Order Meta-Learning Algorithms. https:\/\/arXiv:1803.02999"},{"key":"e_1_3_2_35_2","unstructured":"Emil Njor Jan Madsen and Xenofon Fafoutis. 2023. Data Aware Neural Architecture Search. Retrieved from https:\/\/arxiv:2304.01821"},{"key":"e_1_3_2_36_2","unstructured":"Seongmin Park Beomseok Kwon Jieun Lim Kyuyoung Sim Tae-Ho Kim and Jungwook Choi. 2023. Automatic Network Adaptation for Ultra-Low Uniform-Precision Quantization. Retrieved from https:\/\/arxiv:2212.10878"},{"key":"e_1_3_2_37_2","unstructured":"Archit Parnami and Minwoo Lee. 2022. Learning from Few Examples: A Summary of Approaches to Few-Shot Learning. Retrieved from https:\/\/arxiv:2203.04291"},{"key":"e_1_3_2_38_2","doi-asserted-by":"publisher","DOI":"10.1109\/LSENS.2023.3307119"},{"key":"e_1_3_2_39_2","doi-asserted-by":"publisher","DOI":"10.1145\/3583683"},{"key":"e_1_3_2_40_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.jksuci.2021.11.019"},{"key":"e_1_3_2_41_2","unstructured":"Haoyu Ren Darko Anicic and Thomas Runkler. 2022. How to Manage Tiny Machine Learning at Scale: An Industrial Perspective. Retrieved from https:\/\/arxiv:2202.09113"},{"key":"e_1_3_2_42_2","doi-asserted-by":"publisher","DOI":"10.1109\/IJCNN52387.2021.9533927"},{"key":"e_1_3_2_43_2","doi-asserted-by":"publisher","DOI":"10.1145\/3510820"},{"key":"e_1_3_2_44_2","doi-asserted-by":"publisher","unstructured":"Haoyu Ren Darko Anicic and Thomas A. Runkler. 2023. TinyReptile: TinyML with Federated Meta-Learning. Retrieved from https:\/\/arxiv:2304.05201DOI:10.1109\/ijcnn54540.2023.10191845","DOI":"10.1109\/ijcnn54540.2023.10191845"},{"key":"e_1_3_2_45_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-19433-7_48"},{"key":"e_1_3_2_46_2","doi-asserted-by":"crossref","unstructured":"Haoyu Ren Xue Li Darko Anicic and Thomas A. Runkler. 2023. TinyMetaFed: Efficient Federated Meta-Learning for TinyML. Retrieved from https:\/\/arxiv:2307.06822","DOI":"10.1109\/IJCNN54540.2023.10191845"},{"key":"e_1_3_2_47_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-88361-4_38"},{"key":"e_1_3_2_48_2","unstructured":"Marcus R\u00fcb Daniel Maier Daniel Mueller-Gritschneder and Axel Sikora. 2023. TinyProp\u2014Adaptive Sparse Backpropagation for Efficient TinyML On-device Learning. Retrieved from https:\/\/arxiv:2308.09201"},{"key":"e_1_3_2_49_2","first-page":"163","volume-title":"Proceedings of the 17th Internal Semantic Web Conference (ISWC\u201918)","author":"Sagar Samya","year":"2018","unstructured":"Samya Sagar, Maxime Lefran\u00e7ois, Issam Reba\u00ef, Maha Khemaja, Serge Garlatti, Jamel Feki, and Lionel M\u00e9dini. 2018. Modeling smart sensors on top of SOSA\/SSN and WoT TD with the semantic smart sensor network (S3N) modular ontology. In Proceedings of the 17th Internal Semantic Web Conference (ISWC\u201918). Springer, 163\u2013177."},{"key":"e_1_3_2_50_2","unstructured":"Rafael Stahl Daniel Mueller-Gritschneder and Ulf Schlichtmann. 2023. Fused Depthwise Tiling for Memory Optimization in TinyML Deep Neural Network Inference. Retrieved from https:\/\/arxiv:2303.17878"},{"key":"e_1_3_2_51_2","doi-asserted-by":"publisher","DOI":"10.1109\/SWC50871.2021.00023"},{"key":"e_1_3_2_52_2","first-page":"1","volume-title":"Proceedings of the 60th ACM\/IEEE Design Automation Conference (DAC\u201923)","author":"Delm Josse Van","year":"2023","unstructured":"Josse Van Delm, Maarten Vandersteegen, Alessio Burrello, Giuseppe Maria Sarda, Francesco Conti, Daniele Jahier Pagliari, Luca Benini, and Marian Verhelst. 2023. HTVM: Efficient neural network deployment on heterogeneous TinyML platforms. In Proceedings of the 60th ACM\/IEEE Design Automation Conference (DAC\u201923). IEEE, 1\u20136."},{"key":"e_1_3_2_53_2","doi-asserted-by":"publisher","DOI":"10.1145\/2939502.2939516"},{"key":"e_1_3_2_54_2","doi-asserted-by":"publisher","unstructured":"Pete Warden. 2018. Speech Commands: A Dataset for Limited-Vocabulary Speech Recognition. Retrieved from https:\/\/arxiv.org\/abs\/1804.03209. DOI:10.48550\/ARXIV.1804.03209","DOI":"10.48550\/ARXIV.1804.03209"},{"key":"e_1_3_2_55_2","doi-asserted-by":"publisher","DOI":"10.1145\/3576842.3582382"}],"container-title":["ACM Transactions on Embedded Computing Systems"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3665278","content-type":"unspecified","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3665278","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,6,19]],"date-time":"2025-06-19T00:58:34Z","timestamp":1750294714000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3665278"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2024,6,10]]},"references-count":54,"journal-issue":{"issue":"4","published-print":{"date-parts":[[2024,7,31]]}},"alternative-id":["10.1145\/3665278"],"URL":"https:\/\/doi.org\/10.1145\/3665278","relation":{},"ISSN":["1539-9087","1558-3465"],"issn-type":[{"value":"1539-9087","type":"print"},{"value":"1558-3465","type":"electronic"}],"subject":[],"published":{"date-parts":[[2024,6,10]]},"assertion":[{"value":"2023-12-15","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2024-05-07","order":1,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2024-06-10","order":2,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}