BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//132.216.98.100//NONSGML kigkonsult.se iCalcreator 2.20.4//
BEGIN:VEVENT
UID:20250812T095156EDT-1068tZbc1K@132.216.98.100
DTSTAMP:20250812T135156Z
DESCRIPTION:Title: Invertible Neural Networks - Understanding and Controlli
 ng Learned Representations.\n\nabstract:\n\nOne way to understand deep net
 works is to analyze the information they discard about the input from laye
 r to layer. However\, estimating mutual information between input and hidd
 en representations is intractable in high dimensional problems. Invertible
  deep networks circumvent this problem by guaranteeing information preserv
 ation. In this talk\, I will discuss surprising similarities between non-i
 nvertible and invertible deep networks. Further\, I will discuss how inver
 tible models give rise to an alternative viewpoint on adversarial examples
 . Under this viewpoint adversarial examples are a consequence of excessive
  invariances learned by the classifier\, manifesting themselves in strikin
 g failures when evaluating the model on out of distribution inputs. I will
  discuss how the commonly used cross-entropy objective encourages such ove
 rly invariant representations. Finally\, I will present an extension to cr
 oss-entropy that\, by exploiting properties of invertible deep networks\, 
 enables control of erroneous invariances in theory and practice.\n
DTSTART:20191125T210000Z
DTEND:20191125T220000Z
LOCATION:Room 1104\, Burnside Hall\, CA\, QC\, Montreal\, H3A 0B9\, 805 rue
  Sherbrooke Ouest
SUMMARY:Jörn Jacobsen (University of  Gothenburg)
URL:/mathstat/channels/event/jorn-jacobsen-university-
 gothenburg-302756
END:VEVENT
END:VCALENDAR