CINXE.COM

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">     <html xmlns="http://www.w3.org/1999/xhtml" lang="en" xml:lang="en" xmlns:og="http://ogp.me/ns#" xmlns:fb="http://ogp.me/ns/fb#"> <head> <title>An introduction to Bayesian neural networks and uncertainty quantification in neural networks - CERN Document Server</title> <link href='https://framework.web.cern.ch/framework/2.0/fonts/PTSansWeb/PTSansWeb.css' rel='stylesheet' type='text/css' /> <link rel="stylesheet" href="https://cds.cern.ch/img/invenio.css?v=20141127" type="text/css" /> <link rel="stylesheet" href="https://cds.cern.ch/img/cern_theme/css/cern_theme.css?v=20141127" type="text/css" /> <link rel="stylesheet"href="/css/font-awesome.min.css"> <meta http-equiv="X-UA-Compatible" content="IE=Edge"/> <link rel="stylesheet" href="https://cds.cern.ch/img/cern_toolbar/css/toolbar.css" type="text/css" />    <link rel="canonical" href="https://cds.cern.ch/record/2852107/export/hm" /> <link rel="alternate" hreflang="el" href="https://cds.cern.ch/record/2852107/export/hm?ln=el" /> <link rel="alternate" hreflang="fr" href="https://cds.cern.ch/record/2852107/export/hm?ln=fr" /> <link rel="alternate" hreflang="bg" href="https://cds.cern.ch/record/2852107/export/hm?ln=bg" /> <link rel="alternate" hreflang="zh-TW" href="https://cds.cern.ch/record/2852107/export/hm?ln=zh_TW" /> <link rel="alternate" hreflang="pt" href="https://cds.cern.ch/record/2852107/export/hm?ln=pt" /> <link rel="alternate" hreflang="no" href="https://cds.cern.ch/record/2852107/export/hm?ln=no" /> <link rel="alternate" hreflang="hr" href="https://cds.cern.ch/record/2852107/export/hm?ln=hr" /> <link rel="alternate" hreflang="ca" href="https://cds.cern.ch/record/2852107/export/hm?ln=ca" /> <link rel="alternate" hreflang="de" href="https://cds.cern.ch/record/2852107/export/hm?ln=de" /> <link rel="alternate" hreflang="it" href="https://cds.cern.ch/record/2852107/export/hm?ln=it" /> <link rel="alternate" hreflang="zh-CN" href="https://cds.cern.ch/record/2852107/export/hm?ln=zh_CN" /> <link rel="alternate" hreflang="sv" href="https://cds.cern.ch/record/2852107/export/hm?ln=sv" /> <link rel="alternate" hreflang="sk" href="https://cds.cern.ch/record/2852107/export/hm?ln=sk" /> <link rel="alternate" hreflang="en" href="https://cds.cern.ch/record/2852107/export/hm?ln=en" /> <link rel="alternate" hreflang="pl" href="https://cds.cern.ch/record/2852107/export/hm?ln=pl" /> <link rel="alternate" hreflang="ru" href="https://cds.cern.ch/record/2852107/export/hm?ln=ru" /> <link rel="alternate" hreflang="ka" href="https://cds.cern.ch/record/2852107/export/hm?ln=ka" /> <link rel="alternate" hreflang="ja" href="https://cds.cern.ch/record/2852107/export/hm?ln=ja" /> <link rel="alternate" hreflang="es" href="https://cds.cern.ch/record/2852107/export/hm?ln=es" /> <link rel="alternate" type="application/rss+xml" title="CERN Document Server RSS" href="/rss?ln=en" /> <link rel="search" type="application/opensearchdescription+xml" href="https://cds.cern.ch/opensearchdescription" title="CERN Document Server" /> <link rel="unapi-server" type="application/xml" title="unAPI" href="https://cds.cern.ch/unapi" /> <link rel="apple-touch-icon" href="/apple-touch-icon.png"/> <link rel="apple-touch-icon-precomposed" href="/apple-touch-icon-precomposed.png"/> <meta http-equiv="Content-Type" content="text/html; charset=utf-8" /> <meta http-equiv="Content-Language" content="en" /> <meta name="description" content="Over the past few years, many advances in the field of Deep Learning (DL) have been achieved and nowadays modern DL models are starting to be deployed in our everyday life. However, for many safety-critical applications, as long as scientific research fields, the quantification of the uncertainty of DL model predictions plays a crucial role. In this lecture, I will introduce the basics of Bayesian Neural Networks, how they can tackle the problem of estimating model uncertainty, and the most common techniques for generalizing this method to deep neural networks. Talpini, Jacopo" /> <meta name="keywords" content="CERN Document Server, WebSearch, CERN Document Server" /> <script type="text/javascript" src="https://cds.cern.ch/js/jquery.min.js"></script>  <link rel="stylesheet" href="https://cds.cern.ch/img/webnews.css" type="text/css" />  <script type="text/javascript" src="https://cds.cern.ch/js/webnews.js?v=20131009"></script> <meta property="fb:app_id" content="137353533001720"/>  <meta content="An introduction to Bayesian neural networks and uncertainty quantification in neural networks" name="citation_title" /> <meta content="2023/03/09" name="citation_publication_date" /> <meta name="citation_online_date" content="2023/03/10">  <meta content="An introduction to Bayesian neural networks and uncertainty quantification in neural networks" property="og:title" /> <meta content="Inverted CERN School of Computing 2023" property="og:title" /> <meta content="website" property="og:type" /> <meta content="website" property="og:type" /> <meta content="https://cds.cern.ch/record/2852107" property="og:url" /> <meta property="og:video:height" content="360" /> <meta property="og:video:width" content="640" /> <meta property="og:video" content="https://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presentation-480p-quality.mp4" /> <meta property="og:video:type" content="video/mp4" /> <meta property="og:image" content="https://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presentation-cover.jpg" /> <meta name="twitter:player:height" content="360" /> <meta name="twitter:player:width" content="640" /> <link rel="image_src" href="https://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presentation-cover.jpg" /> <link rel="video_src" href="https://cds.cern.ch/mediaplayer.swf?file=/1208723c37/1208723c37-presentation-480p-quality.mp4&streamer=rtmp://wowza.cern.ch:1935/vod&provider=rtmp&stretching=exactfit&image=https://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presentation-cover.jpg"/> <meta name="twitter:player" content="https://cds.cern.ch/video/?"/> <meta content="CERN Document Server" property="og:site_name" /> <meta content="Over the past few years, many advances in the field of Deep Learning (DL) have been achieved and nowadays modern DL models are starting to be deployed in our everyday life. However, for many safety-critical applications, as long as scientific research fields, the quantification of the uncertainty of DL model predictions plays a crucial role. In this lecture, I will introduce the basics of Bayesian Neural Networks, how they can tackle the problem of estimating model uncertainty, and the most common techniques for generalizing this method to deep neural networks." property="og:description" />  <meta content="summary" name="twitter:card" /> <style></style> </head> <body class="CERN32Document32Server search" lang="en">  <div id="cern-toolbar"> <h1><a href="http://cern.ch" title="CERN">CERN <span>Accelerating science</span></a></h1> <ul> <li class="cern-accountlinks"><a class="cern-account" href="https://cds.cern.ch/youraccount/login?ln=en&referer=https%3A//cds.cern.ch/record/2852107/export/hm%3Fln%3Den" title="Sign in to your CERN account">Sign in</a></li> <li><a class="cern-directory" href="http://cern.ch/directory" title="Search CERN resources and browse the directory">Directory</a></li> </ul> </div>   <div role="banner" class="clearfix" id="header"> <div class="header-inner inner"> <hgroup class="clearfix"> <h2 id="site-name"> <a rel="home" title="Home" href="/"><span>CERN Document Server</span></a> </h2> <h3 id="site-slogan">Access articles, reports and multimedia content in HEP</h3> </hgroup> <div role="navigation" id="main-navigation" class="cdsmenu"> <h2 class="element-invisible">Main menu</h2><ul class="links inline clearfix"> <li class="menu-386 first active-trail"><a class="active-trail" href="https://cds.cern.ch/?ln=en">Search</a></li> <li class="menu-444 "><a class="" title="" href="https://cds.cern.ch/submit?ln=en">Submit</a></li> <li class="menu-426 "><a class="" href="https://cds.cern.ch/help/?ln=en">Help</a></li> <li class="leaf hassubcdsmenu"> <a hreflang="en" class="header" href="https://cds.cern.ch/youraccount/display?ln=en">Personalize</a> <ul class="subsubcdsmenu"><li><a href="https://cds.cern.ch/youralerts/list?ln=en">Your alerts</a></li><li><a href="https://cds.cern.ch/yourbaskets/display?ln=en">Your baskets</a></li><li><a href="https://cds.cern.ch/yourcomments?ln=en">Your comments</a></li><li><a href="https://cds.cern.ch/youralerts/display?ln=en">Your searches</a></li></ul></li> </ul> </div> </div> </div>  <table class="navtrailbox"> <tr> <td class="navtrailboxbody"> <a href="/?ln=en" class="navtrail">Home</a> > <a class="navtrail" href="/record/2852107">An introduction to Bayesian neural networks and uncertainty quantification in neural networks</a> > HTML MARC </td> </tr> </table> </div> <div class="pagebody"><div class="pagebodystripemiddle"> <pre style="margin: 1em 0px;">002852107 001__ 2852107 002852107 005__ 20240626130612.0 002852107 0248_ $$aoai:cds.cern.ch:2852107$$pcerncds:TALK 002852107 041__ $$aeng 002852107 084__ $$2Indico$$a2987 002852107 110__ $$aCERN. Geneva 002852107 111__ $$92023-03-09T10:00:00$$aInverted CERN School of Computing 2023$$cCERN - 31/3-004 - IT Amphitheatre$$g1208723c37$$z2023-03-09T11:00:00 002852107 245__ $$aAn introduction to Bayesian neural networks and uncertainty quantification in neural networks 002852107 260__ $$c2023 002852107 269__ $$c2023-03-09 002852107 300__ $$a2836 002852107 340__ $$aStreaming video 002852107 490__ $$aInverted CSC 002852107 490__ $$aInverted CERN School of Computing 2023 002852107 518__ $$d2023-03-09T10:00:00 002852107 520__ $$aOver the past few years, many advances in the field of Deep Learning (DL) have been achieved and nowadays modern DL models are starting to be deployed in our everyday life. However, for many safety-critical applications, as long as scientific research fields, the quantification of the uncertainty of DL model predictions plays a crucial role. In this lecture, I will introduce the basics of Bayesian Neural Networks, how they can tackle the problem of estimating model uncertainty, and the most common techniques for generalizing this method to deep neural networks. 002852107 542__ $$dCERN$$g2023 002852107 65017 $$aInverted CSC 002852107 65027 $$aEvent 002852107 690C_ $$aTALK 002852107 690C_ $$aCERN 002852107 700__ $$aTalpini, Jacopo$$espeaker$$uUniversità degli Studi di Milano-Bicocca 002852107 859__ $$fsebastian.lopienski@cern.ch 002852107 8564_ $$uhttps://indico.cern.ch/event/1208723/contributions/5230073/$$yTalk details 002852107 8564_ $$uhttps://indico.cern.ch/event/1208723/$$yEvent details 002852107 8567_ $$2MediaArchive$$d/mnt/master_share/master_data/2023/1208723c37$$xAbsolute master path 002852107 8567_ $$2MediaArchive$$uhttps://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presentation-cover.jpg$$xpngthumbnail$$ythumbnail weblecture 002852107 8567_ $$2MediaArchive$$uhttps://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presentation-480p-quality.mp4$$xvideo/mp4$$yContent: presentation. Resolution: 640x360. Baudrate: 200990 002852107 8567_ $$2MediaArchive$$uhttps://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presentation-1080p-quality.mp4$$xvideo/mp4$$yContent: presentation. Resolution: 1920x1080. Baudrate: 831903 002852107 8567_ $$2MediaArchive$$uhttps://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presentation-720p-quality.mp4$$xvideo/mp4$$yContent: presentation. Resolution: 1280x720. Baudrate: 503560 002852107 8567_ $$2MediaArchive$$uhttps://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presentation-360p-quality.mp4$$xvideo/mp4$$yContent: presentation. Resolution: 480x270. Baudrate: 93341 002852107 8567_ $$2MediaArchive$$uhttps://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presenter-480p-quality.mp4$$xvideo/mp4$$yContent: presenter. Resolution: 640x360. Baudrate: 270715 002852107 8567_ $$2MediaArchive$$uhttps://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presenter-720p-quality.mp4$$xvideo/mp4$$yContent: presenter. Resolution: 1280x720. Baudrate: 902372 002852107 8567_ $$2MediaArchive$$uhttps://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presenter-1080p-quality.mp4$$xvideo/mp4$$yContent: presenter. Resolution: 1920x1080. Baudrate: 2444113 002852107 8567_ $$2MediaArchive$$uhttps://lecturemedia.cern.ch/2023/1208723c37/1208723c37-presenter-360p-quality.mp4$$xvideo/mp4$$yContent: presenter. Resolution: 480x270. Baudrate: 85643 002852107 906__ $$pPace, Alberto$$uCERN 002852107 906__ $$pGunne, Kristina$$uCERN 002852107 906__ $$pPolok, Jarek$$uCERN 002852107 961__ $$c2023-03-10T12:33:14$$x2022-10-05T07:53:46 002852107 963__ $$aPUBLIC 002852107 970__ $$aINDICO.1208723c37 002852107 980__ $$aIndico$$bSSW</pre></div></div> <footer id="footer" class="pagefooter clearfix">  <div class="pagefooterstripeleft"> CERN Document Server :: <a class="footer" href="https://cds.cern.ch/?ln=en">Search</a> :: <a class="footer" href="https://cds.cern.ch/submit?ln=en">Submit</a> :: <a class="footer" href="https://cds.cern.ch/youraccount/display?ln=en">Personalize</a> :: <a class="footer" href="https://cds.cern.ch/help/?ln=en">Help</a> :: <a class="footer" href="https://cern.service-now.com/service-portal?id=privacy_policy&se=CDS-Service" target="_blank">Privacy Notice</a> :: <a class="footer" href="https://repository.cern/content-policy" target="_blank">Content Policy</a> :: <a class="footer" href="https://repository.cern/terms" target="_blank">Terms and Conditions</a> <br /> Powered by <a class="footer" href="http://invenio-software.org/">Invenio</a> <br /> Maintained by <a class="footer" href="https://cern.service-now.com/service-portal?id=service_element&name=CDS-Service">CDS Service</a> - Need help? Contact <a href="https://cern.service-now.com/service-portal?id=service_element&name=CDS-Service">CDS Support</a>. <br /> </div> <div class="pagefooterstriperight"> <div class="cern-logo"> <a id="logo" href="http://cern.ch" title="CERN" rel="CERN" ><img src="https://cds.cern.ch/img/cern_theme/img/cern-logo-large.png" alt="CERN" /></a> </div> <div class="cern-languagebox"> This site is also available in the following languages:<br /><a href="/record/2852107/export/hm?ln=bg" class="langinfo">Български</a>  <a href="/record/2852107/export/hm?ln=ca" class="langinfo">Català</a>  <a href="/record/2852107/export/hm?ln=de" class="langinfo">Deutsch</a>  <a href="/record/2852107/export/hm?ln=el" class="langinfo">Ελληνικά</a>  <span class="langinfo">English</span>  <a href="/record/2852107/export/hm?ln=es" class="langinfo">Español</a>  <a href="/record/2852107/export/hm?ln=fr" class="langinfo">Français</a>  <a href="/record/2852107/export/hm?ln=hr" class="langinfo">Hrvatski</a>  <a href="/record/2852107/export/hm?ln=it" class="langinfo">Italiano</a>  <a href="/record/2852107/export/hm?ln=ja" class="langinfo">日本語</a>  <a href="/record/2852107/export/hm?ln=ka" class="langinfo">ქართული</a>  <a href="/record/2852107/export/hm?ln=no" class="langinfo">Norsk/Bokmål</a>  <a href="/record/2852107/export/hm?ln=pl" class="langinfo">Polski</a>  <a href="/record/2852107/export/hm?ln=pt" class="langinfo">Português</a>  <a href="/record/2852107/export/hm?ln=ru" class="langinfo">Русский</a>  <a href="/record/2852107/export/hm?ln=sk" class="langinfo">Slovensky</a>  <a href="/record/2852107/export/hm?ln=sv" class="langinfo">Svenska</a>  <a href="/record/2852107/export/hm?ln=zh_CN" class="langinfo">中文(简)</a>  <a href="/record/2852107/export/hm?ln=zh_TW" class="langinfo">中文(繁)</a> </div> </div>  </footer> <script type="text/javascript"> var SyndeticsBookCovers = (function() { var SMALL_SIZE = "sc.gif", MEDIUM_SIZE = "mc.gif", RAW_URL = "https://secure.syndetics.com/index.aspx?isbn=THEISBN/THESIZE&client=cernlibrary"; replaceCover = function(imgElement, isbns, hdFormat) { var img = new Image(), size = hdFormat ? MEDIUM_SIZE : SMALL_SIZE; var _isbns = isbns.sort(function(a, b) { // sort from shortest to longest ISBN (more modern) return a.length > b.length ? 1 : -1; }); function next() { var isbn = _isbns.pop(); if (isbn) { var url = RAW_URL.replace("THEISBN", isbn).replace("THESIZE", size); img.src = url; } } function done() { imgElement.src = img.src; } img.onload = function() { if (this.width > 1) { done(); } else { next(); } }; next(); }; return { replaceCover: replaceCover }; })(); $(document).ready(function() { // get book covers $("img.book-cover").each(function() { var $this = $(this), strIsbns = $this.data("isbns") || "", isbnsArray = String(strIsbns).split(","), hdFormat = $this.hasClass("hd"); SyndeticsBookCovers.replaceCover(this, isbnsArray, hdFormat); }); // WebNews tooltips $.ajax({ url: "/news/tooltips", success: function(data) { create_tooltips(data); }, dataType: "json", cache: false }); }); </script>  <script src="//cds.cern.ch/js/feedback.js"></script>   <script> var _paq = window._paq = window._paq || []; /* tracker methods like "setCustomDimension" should be called before "trackPageView" */ _paq.push(['trackPageView']); _paq.push(['enableLinkTracking']); (function() { var u="https://webanalytics.web.cern.ch/"; _paq.push(['setTrackerUrl', u+'matomo.php']); _paq.push(['setSiteId', '756']); var d=document, g=d.createElement('script'), s=d.getElementsByTagName('script')[0]; g.async=true; g.src=u+'matomo.js'; s.parentNode.insertBefore(g,s); })(); </script>  </body> </html>