BEGIN:VCALENDAR VERSION:2.0 PRODID:-//wp-events-plugin.com//7.4.0.1//EN TZID:Europe/Paris X-WR-TIMEZONE:Europe/Paris BEGIN:VEVENT UID:311@isdm.umontpellier.fr DTSTART;TZID=Europe/Paris:20230420T140000 DTEND;TZID=Europe/Paris:20230420T140000 DTSTAMP:20260828T080902Z URL:https://isdm.umontpellier.fr/events/stochastic-local-winner-takes-all- networks/ SUMMARY:Stochastic Local Winner-Takes-All Networks DESCRIPTION:Room 109\, Building 9\, St Eloi campus\n\nMachine Learning in M ontpellier\, Theory &\; Practice\n\nThe recent mass adoption of DNNs\, even in safety-critical scenarios\, has shifted the focus of the research community towards the creation of inherently intrepretable models. Concept Bottleneck Models (CBMs) constitute a popular approach where hidden layer s are tied to human understandable concepts allowing for investigation and correction of the network'\;s decisions. However\, CBMs however usual ly suffer from: (i) performance degradation and (ii) lower interpretabilit y than intended due to the sheer amount of concepts contributing to each d ecision. In this work\, we propose a simple yet highly intuitive interpret able framework based on Contrastive Language Image models and a single spa rse linear layer. In stark contrast to related approaches\, the sparsity i n our framework is achieved via principled Bayesian arguments by inferring concept presence via a data-driven Bernoulli distribution. As we experime ntally show\, our framework not only outperforms recent CBM approaches acc uracy-wise\, but it also yields high per example concept sparsity\, facili tating the individual investigation of the emerging concepts.\,\, |\n\nMac hine Learning in Montpellier\, Theory &\; Practice ATTACH;FMTTYPE=image/jpeg:https://isdm.umontpellier.fr/wp-content/uploads/ 2026/06/ml-mtp-gC78d5.png CATEGORIES:ML MTP LOCATION:Saint Eloi Campus\, Building 9\, Room 109\, Montpellier\, X-APPLE-STRUCTURED-LOCATION;VALUE=URI;X-ADDRESS=Montpellier\, ;X-APPLE-RADI US=100;X-TITLE=Saint Eloi Campus\, Building 9\, Room 109:geo:0,0 END:VEVENT BEGIN:VTIMEZONE TZID:Europe/Paris X-LIC-LOCATION:Europe/Paris BEGIN:DAYLIGHT DTSTART:20230326T030000 TZOFFSETFROM:+0100 TZOFFSETTO:+0200 TZNAME:CEST END:DAYLIGHT END:VTIMEZONE END:VCALENDAR