[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-119986-en":3,"doc-seo-119986-105":30,"detail-sidebar-cat-0-en-105":92},{"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":27,"seo_description":14,"update_tm":28,"read_time":29},119986,549758252649,"Ivy","https://ap-avatar.wpscdn.com/avatar/8000253669c5317157?_k=1778319167496531819",8,"Research & Report","Uncertainty Estimation: Single Forward Pass Methods and Applications in Active Learning - Doctor of Philosophy thesis","Machine Learning (ML) models are now widely used in high-stakes automated decision-making, yet they can still make errors. Reliable deployment requires quantifying predictive uncertainty and incorporating it into user decisions. Deep learning models often struggle to express uncertainty directly, limiting safety in real-world settings. Bayesian approaches provide a principled framework, but exact inference is computationally costly and approximations can compromise accuracy or remain expensive. The thesis analyzes this trade-off and proposes uncertainty estimation models that perform well using a single forward pass within active learning applications.","Uncertainty Estimation: Single Forward Pass Methods and Applications in Active Learning  \nJoost René van Amersfoort  \nWolfson College University of Oxford  \nA thesis submitted for the degree of Doctor of Philosophy  \nTrinity 2022  \nAcknowledgements  \nI am very fortunate to have been part of two great labs: OATML and OxCSML. When I joined OATML as part of its first few DPhils students, we had the opportunity to craft a new lab that we truly enjoined. We tried out many talk formats, compute setups, collaborations, office arrangements, and research directions. I have learned so much building all of this up with the rest of OATML, and I am very happy that I was there for the beginning. OxCSML provided a welcome stable environment, with tried and tested talk formats and long-standing research directions. A comforting alternative to OATML’s constant change.  \nYarin, I am impressed with what you have built up with OATML over just a few years. Thank you for taking me along on this ride. Our many conversations were always a source of energy and inspiration. My favourite moments were when I was convinced your reply did not make sense. To only later realise, that you did understand me and were already some steps ahead.  \nYee Whye, I continue to be amazed by the number of things you are involved in. You manage to have an impact on such a wide array of projects, with the special highlight being COVID-19 modelling at SAGE. Your breadth of knowledge is a real inspiration, and somehow you always ask just the right question at the right moment. At OATML, no one had more impact on my research than Lewis Smith. Our conversations have made this thesis possible. Sebastian Farquhar, Milad Alizadeh, Andreas Kirsch, Andrew Jesson, Jishnu Mukhoti, Panagiotis Tigas, Pascal Notin, Aidan Gomez, Oscar Key, Chetan Gohil, Bas Veeling, Adam Foster, Emilien Dupont, Haiwen Huang, and John Ryan: I had a lot of fun working with each and everyone of you. You all taught me something different, and I am curious to see where you will go. Angelos Filos, Clare Lyle, Freddie Bickford Smith, Gunshi Gupta, Jan Brauner, Jannik Kossen, Kelsey Doerksen, Lisa Schut, Lorenz Kuhn, Sören Mindermann, and Tim Rudner: thank you for the fun discussions and making my time in Oxford so enjoyable. A special thank you to Ian Collier of the IT team atthe Computer Science department–your work has made a big difference. I hope OATML remains the productive and high energy place for a long time to come!  \nKorfball was the heart of my non-academic life at Oxford. Thank you to Sophie, Joe, Matt, Amy, Stephan, Niall and all other Korf friends for providing many hours of fun, four won varsities, and one league championship. Kara Allum, you have influenced my view on life more than you know. I hope that we remain friends for a long time.  \nGratitude to the following Oxford establishments from one happy customer: ATS, Najar’s, Aleppo’s, Opera, Hamblin, Tse Noodle, Za’atar, Zhang Ji, The Chester Arms, Oli’s Thai, Peppers, The Greek Takeaway, Currydor, Brew, Missing Bean, 101 Coffee, The Victoria, The Star, The Cape of Good Hope, The Fir Tree, and The Royal Oak.  \nMy DPhil would not have been the same, or even materialised really, without Luisa Zintgraf. Your help and encouragement have made the difference during so many moments. We survived moving to a new country, a pandemic, and two DPhils, I am sure we are ready for everything that comes next. Thank you for being my source of chaos and love.  \nI want to thank my family for all their support, (surprise) visits, packages, and phone calls. My sisters Emma and Irene, my parents Wilma and Steef, you are all wonderful. Thank you to my extended family Chiara, Gabriele, Deborah, and Reinhold, I look forward to hosting you many more times.  \nKyriacos, Eva, Jonas and Panos, thank you for keeping me cool and fun in Oxford. To my Dutch(-based) friends, Thijmen, Lukas, Pascal, Otto, Yvonne, Hüseyin, Bas, Thijs, Marcel, Joost–“Doorn”, thank you for support a","cbCaiojvE25tuvMH","https://ap.wps.com/l/cbCaiojvE25tuvMH","pdf",15017543,1,147,"English","en",105,"# Abstract\n## Background and motivation\n## Core approach and desiderata\n## DUQ model for single forward-pass uncertainty\n## Extensions via Deep Kernel Learning","[{\"question\":\"Why is uncertainty quantification important for machine learning in real-world applications?\",\"answer\":\"Even accurate ML models can still make mistakes, so dependable use requires expressing how uncertain the predictions are and ensuring users take that uncertainty into account.\"},{\"question\":\"What limitation motivates the thesis beyond standard deep learning models?\",\"answer\":\"Deep learning models cannot readily express their uncertainty, which makes them unsafe for many real-world applications where decision reliability is critical.\"},{\"question\":\"How does the thesis achieve uncertainty estimation using a single forward pass?\",\"answer\":\"It introduces a model (DUQ) that estimates uncertainty in one forward pass by carefully constructing the model’s parameter and output space according to specified desiderata.\"}]","Uncertainty Estimation: Single Forward Pass Methods and Applications in Active Learning - Doctor of Philosophy thesis | PDF",1785727496,370,{"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":87,"head_meta":89,"extra_data":91,"updated_unix":28},"uncertainty-estimation-single-forward-pass-methods-and-applications-in-active-learning-doctor-of-philosophy-thesis","",{"@graph":36,"@context":86},[37,54,69],{"@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/uncertainty-estimation-single-forward-pass-methods-and-applications-in-active-learning-doctor-of-philosophy-thesis/119986/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-04","2026-08-03",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"Why is uncertainty quantification important for machine learning in real-world applications?","Question",{"text":76,"@type":77},"Even accurate ML models can still make mistakes, so dependable use requires expressing how uncertain the predictions are and ensuring users take that uncertainty into account.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What limitation motivates the thesis beyond standard deep learning models?",{"text":81,"@type":77},"Deep learning models cannot readily express their uncertainty, which makes them unsafe for many real-world applications where decision reliability is critical.",{"name":83,"@type":74,"acceptedAnswer":84},"How does the thesis achieve uncertainty estimation using a single forward pass?",{"text":85,"@type":77},"It introduces a model (DUQ) that estimates uncertainty in one forward pass by carefully constructing the model’s parameter and output space according to specified desiderata.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":93},[94,98,102,106,111,116,121,124,129,132,136],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":112,"doc_module":4,"doc_module_name":46,"category_name":113,"show_sort_weight":114,"slug":115},6,"Technology",50,"technology",{"id":117,"doc_module":4,"doc_module_name":46,"category_name":118,"show_sort_weight":119,"slug":120},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":122,"slug":123},30,"research-report",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":126,"show_sort_weight":127,"slug":128},9,"Religion & Spirituality",20,"religion-spirituality",{"id":127,"doc_module":4,"doc_module_name":46,"category_name":130,"show_sort_weight":127,"slug":131},"World Cup","world-cup",{"id":133,"doc_module":4,"doc_module_name":46,"category_name":134,"show_sort_weight":133,"slug":135},10,"Lifestyle","lifestyle",{"id":137,"doc_module":4,"doc_module_name":46,"category_name":138,"show_sort_weight":107,"slug":139},19,"General","general"]