BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:#optazur Exact Continuous Relaxations of L0-Regularized Generalize
 d Linear Models
DTSTART:20240603T120000Z
DTEND:20240603T130000Z
DTSTAMP:20260809T201500Z
UID:indico-event-10949@indico.math.cnrs.fr
DESCRIPTION:Speakers: Emmanuel Soubies\n\nSparse generalized linear models
  are widely used in fields such as statistics\, computer vision\, signal/i
 mage processing and machine learning. The natural sparsity promoting regul
 arizer is the l0 pseudo-norm which is discontinuous and non-convex. In thi
 s talk\, we will present the l0-Bregman relaxation (B-Rex)\, a general fra
 mework to compute exact continuous relaxations of such l0-regularized crit
 eria. Although in general still non-convex\, these continuous relaxations 
 are qualified as exact in the sense that they let unchanged the set of glo
 bal minimizer while enjoying a better optimization landscape. In particula
 r\, we will show that some local minimizers of the initial functional are 
 eliminated by these relaxations. Finally\, these properties will be illust
 rated on both sparse Kullback-Leibler regression and sparse logistic regre
 ssion problems.\n\nhttps://indico.math.cnrs.fr/event/10949/
LOCATION:I3S
URL:https://indico.math.cnrs.fr/event/10949/
END:VEVENT
END:VCALENDAR