[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84413-en":3,"doc-seo-84413-105":29,"detail-sidebar-cat-0-en-105":83},{"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":20,"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":13,"seo_description":14,"update_tm":27,"read_time":28},84413,1099513958607,"Jiven","https://ap-avatar.wpscdn.com/avatar/100002390cf8733938c?x-image-process=image/resize,m_fixed,w_180,h_180&k=1778829742770036399",8,"Research & Report","Neural Active Learning Meets the Partial Monitoring Framework","The document studies online-based active learning (OAL) in which an agent processes a stream of observations and balances expensive label acquisition against the cost of prediction errors. It introduces a theoretical foundation for OAL using partial monitoring, showing that established binary and multi-class OAL formulations are instances of partial monitoring. It further extends OAL toward cost-sensitive settings by proposing NeuralCBP, the first partial monitoring strategy leveraging deep neural networks for predictive uncertainty, evaluated competitively on multiple benchmark datasets.","Neural Active Learning Meets the Partial Monitoring Framework  \nMaxime Heuillet 1,4,5 Ola Ahmad 1,3 Audrey Durand 1,2,4,5  \n1 Université Laval, Canada  \n2 Canada-CIFAR AI Chair  \n3 Thales Research and Technology (cortAIx), Canada,  \n4 Mila-Québec AI Institute, Canada,  \n5 Institut Intelligence et Données, Canada,  \narXiv :2405 .0892 1v2 [ cs .LG] 12 Jul 2026  \nAbstract  \nWe focus on the online-based active learning (OAL) setting where an agent operates over a stream of observations and trades-off between the costly acquisition of information (labelled observations) and the cost of prediction errors. We propose a novel foundation for OAL tasks based on partial monitoring, a theoretical framework specialized in online learning from partially informative actions. We show that previously studied binary and multi-class OAL tasks are instances of partial monitoring.  \nWe expand the real-world potential of OAL by introducing a new class of cost-sensitive OAL tasks. We propose NeuralCBP, the first PM strategy that accounts for predictive uncertainty with deep neural networks. Our extensive empirical evaluation on open source datasets shows that NeuralCBP has competitive performance against state-of-the-art baselines on multiple binary, multi-class and costsensitive OAL tasks.  \n1 INTRODUCTION  \nIn active learning [Cohn et al., 1994], an agent decides to query an expert to obtain labels on selected observations. This active acquisition of labels efficiently reduces the number of labelled observations needed to learn a task. Active learning therefore appears as a costeffective solution for modern machine learning, which often relies on large volumes of labelled observations [Kusne et al., 2020] .  \nIn this work, we focus on the online-based active learning (OAL) setting for binary and multi-class classification tasks Beygelzimer et al. [2009] . The agent operates over a (possibly infinite) stream of observations. For  \neach observation, the agent predicts the class and either decides to reveal its prediction or to query an expert to obtain the label. The OAL setting we consider differs from the batch setting where the agent gathers fixed-size batches of observations to label Saran et al.[2023], Amin et al. [2020] . In both the OAL and the batch-based settings, all decisions are irrevocable and associated with costs. The goal is to minimize the cumulative cost over the stream of decisions, by trading-off between the cost of obtaining new labels (labeling complexity) and the cost of prediction errors (generalization performance) .  \nIn the context of OAL for binary classification, the Margin strategy [Sculley, 2007] queries the expert when the prediction uncertainty is greater than a user-specified threshold. In contrast, with Cesa [CesaBianchi et al., 2006], labelled observations are acquired proportionally to the global prediction error rate of the strategy. Both Margin and Cesa are specifically analyzed for the class of linear separators and are designed for binary tasks. More recent studies focused on multi-class OAL tasks. The Gappletron [van der Hoeven et al., 2021] leverages graph feedback, making it inherently multi-class. However, simialrly to Cesa and Margin, Gappletron is specifically analyzed for linear separators.  \nModern applications of machine learning involve highdimensional observations that require learning complex representations. As a result, Neural [Wang et al., 2021] and ALPS [DeSalvo et al., 2021] proposed multi-class OAL strategies based on deep neural networks. Neural and ALPS have been outperformed by INeural [Ban et al., 2022b], an improved and more practical version of the Neural strategy [Wang et al., 2021] . The current state-of-the-art, Neuronal [Ban et al., 2024], addresses scalability limitations of INeural, opening the door to using sophisticated neural architectures, such as convolutional neural networks.  \nProceedings of the 40th Conference on Uncertainty in Artificial Intelligence (UAI 2024), PMLR 244","cbCairV8LmznUpTc","https://ap.wps.com/l/cbCairV8LmznUpTc","pdf",1325780,1,21,"English","en",105,"# Abstract\n# Introduction\n## Online active learning setting\n## Related work and limitations\n## Problem motivation: cost-sensitive OAL\n## Contributions","[{\"question\":\"What is NeuralCBP and why is it significant?\",\"answer\":\"NeuralCBP is a partial monitoring strategy designed to learn with deep neural networks by accounting for predictive uncertainty; experiments show competitive performance on multiple binary, multi-class, and cost-sensitive OAL tasks.\"}]",1784195464,53,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":78,"head_meta":80,"extra_data":82,"updated_unix":27},"neural-active-learning-meets-the-partial-monitoring-framework","",{"@graph":35,"@context":77},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/neural-active-learning-meets-the-partial-monitoring-framework/84413/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71],{"name":72,"@type":73,"acceptedAnswer":74},"What is NeuralCBP and why is it significant?","Question",{"text":75,"@type":76},"NeuralCBP is a partial monitoring strategy designed to learn with deep neural networks by accounting for predictive uncertainty; experiments show competitive performance on multiple binary, multi-class, and cost-sensitive OAL tasks.","Answer","https://schema.org",{"og:url":51,"og:type":79,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":81,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":84},[85,89,93,97,102,107,112,115,120,123,127],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":86,"show_sort_weight":87,"slug":88},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":90,"show_sort_weight":91,"slug":92},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Exam",70,"exam",{"id":98,"doc_module":4,"doc_module_name":45,"category_name":99,"show_sort_weight":100,"slug":101},5,"Comic",60,"comic",{"id":103,"doc_module":4,"doc_module_name":45,"category_name":104,"show_sort_weight":105,"slug":106},6,"Technology",50,"technology",{"id":108,"doc_module":4,"doc_module_name":45,"category_name":109,"show_sort_weight":110,"slug":111},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":113,"slug":114},30,"research-report",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":45,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":98,"slug":130},19,"General","general"]