[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-124675-en":3,"doc-seo-124675-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},124675,962075006959,"Anda","https://ap-avatar.wpscdn.com/avatar/e0002397efbe92a78e?_k=1776741047341049297",8,"Research & Report","Tiny Machine Learning for Concept Drift - Paper summary","Tiny Machine Learning (TML) aims to run machine and deep learning on embedded systems and IoT devices under strict limits on memory, computation, and energy. Many existing approaches optimize only inference cost and assume training in cloud or edge environments, which can fail when the data-generating process changes. A Tiny Machine Learning for Concept Drift (TML-CD) solution is proposed using deep feature extractors and a k-nearest neighbors classifier with a passive hybrid adaptation module and an active change-detection test. Experiments on image and audio benchmarks, plus deployment on three micro-controller units, validate effectiveness for real-world pervasive systems.","arXiv :2 107 . 14759v 1 [ cs .LG] 30 Jul 2021  \nTiny Machine Learning for Concept Drift  \nSimone Disabato, and Manuel Roveri, Senior Member, IEEE  \nAbstract—Tiny Machine Learning (TML) is a new research area whose goal is to design machine and deep learning techniques able to operate in Embedded Systems and IoT units, hence satisfying the severe technological constraints on memory, computation, and energy characterizing these pervasive devices. Interestingly, the related literature mainly focused on reducing the computational and memory demand of the inference phase of machine and deep learning models. At the same time, the training is typically assumed tobe carried out in Cloud or edge computing systems (due to the larger memory and computational requirements) . This assumption results in TML solutions that might become obsolete when the process generating the data is affected by concept drift (e.g. , due to periodicity or seasonality effect, faults or malfunctioning affecting sensors or actuators, or changes in the users' behavior), a common situation in real-world application scenarios. For the first time in the literature, this paper introduces a Tiny Machine Learning for Concept Drift (TML-CD) solution based on deep learning feature extractors and a k-nearest neighbors classifier integrating a hybrid adaptation module able to deal with concept drift affecting the data-generating process. This adaptation module continuously updates (in a passive way) the knowledge base of TML-CD and, at the same time, employs a Change Detection Test to inspect for changes (inan active way) to quickly adapt to concept drift by removing the obsolete knowledge. Experimental results on both image and audio benchmarks show the effectiveness of the proposed solution, whilst the porting of TML-CD on three off-the-shelf micro-controller units shows the feasibility of what is proposed in real-world pervasive systems.  \nIndex Terms—Tiny Machine Learning, Concept Drift, Adaptation, Deep Learning, k-Nearest Neighbour.  \n1 INTRODUCTION  \nInternet-of-Things (IoT) and embedded systems are nowadays part of our everyday life in a wide range of application scenarios (e.g., automotive, medical devices, and smart cities, to name a few) . In recent years, the scientificand technological trend about these pervasive devices is to move the processing (and in particular the intelligent processing) as close as possible to where data are generated. The reason is twofold. First, IoT units and embedded systems already operate pervasively in the environment processing large amounts of data acquired by the sensors. Second, machine and deep learning solutions processing these data directly on the pervasive devices are crucial to support real-time applications, prolong the system lifetime, and increase the Quality-of-Service. Nevertheless, machine and deep learning solutions are typically characterized by memory and computational demands that rarely match the constraints on memory, computation, and energy characterizing the IoT units and embedded systems [1], [2],[3] .  \nTiny Machine Learning (TML) [4] is a relatively new research area aiming at filling this gap by designing “tiny”machine and deep learning solutions able to run on IoT units and embedded systems. Section 2 analyses the related literature, highlighting that most TML solutions focus on approximation, pruning, and quantization mechanisms to reduce memory and computational demand of machine and deep learning models. Although these solutions run on embedded systems and IoT units, their training is typically carried out on high-performing units (such as Cloud or EdgeComputing systems), with very few papers proposing on-device incremental learning mechanisms [5],[6] .  \n􀀏 S. Disabato and M. Roveri are with the Dipartimento di Elettronica, Informazione e Bioingegneria (DEIB), Politecnico di Milano, Milan, Italy. E-mail: fsimone.disabato,[manuel.roveri](manuel.roverig@polimi.it)[g](manuel.roverig@polimi.it)[@polimi.it]","cbCaifl24fOgvfbv","https://ap.wps.com/l/cbCaifl24fOgvfbv","pdf",1325182,1,11,"English","en",105,"# Introduction\n# Related Literature\n# Problem Formalization\n# Proposed TML-CD Solution\n## Adaptation Mechanisms (Passive, Active, Hybrid)\n## Stages and Deployment\n# Experimental Results\n# Conclusions","[{\"question\":\"Why does concept drift make many TML approaches ineffective?\",\"answer\":\"Because training is often assumed to happen in cloud or edge systems, while real deployments can experience changes in the data-generating process. When drift occurs, models may not update and accuracy can drop over time.\"},{\"question\":\"What is the core idea of the proposed TML-CD solution?\",\"answer\":\"TML-CD integrates deep learning feature extractors with a k-nearest neighbors classifier and a hybrid adaptation module. The module updates knowledge passively and uses a change detection test actively to remove obsolete knowledge.\"},{\"question\":\"How is TML-CD evaluated and validated?\",\"answer\":\"It is tested on both image and audio benchmarks to show effectiveness. It is also ported to three off-the-shelf micro-controller units to demonstrate feasibility in real pervasive systems.\"}]","Tiny Machine Learning for Concept Drift - Paper summary | PDF",1785893856,28,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"tiny-machine-learning-for-concept-drift-paper-summary","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/tiny-machine-learning-for-concept-drift-paper-summary/124675/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-05",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why does concept drift make many TML approaches ineffective?","Question",{"text":75,"@type":76},"Because training is often assumed to happen in cloud or edge systems, while real deployments can experience changes in the data-generating process. When drift occurs, models may not update and accuracy can drop over time.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is the core idea of the proposed TML-CD solution?",{"text":80,"@type":76},"TML-CD integrates deep learning feature extractors with a k-nearest neighbors classifier and a hybrid adaptation module. The module updates knowledge passively and uses a change detection test actively to remove obsolete knowledge.",{"name":82,"@type":73,"acceptedAnswer":83},"How is TML-CD evaluated and validated?",{"text":84,"@type":76},"It is tested on both image and audio benchmarks to show effectiveness. It is also ported to three off-the-shelf micro-controller units to demonstrate feasibility in real pervasive systems.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]