BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Explaining Non-Linear Classifier Decisions with application to Dee
 p Learning - Prof Klaus-Robert Müller (TU Berlin)
DTSTART:20150629T140000Z
DTEND:20150629T150000Z
UID:TALK60039@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:Understanding and interpreting classification decisions of aut
 omated image classification systems is of\nhigh value in many applications
 \, as it allows to verify the reasoning of the system and provides additio
 nal\ninformation to the human expert. Although machine learning methods ar
 e solving very successfully a\nplethora of tasks\, they have in most cases
  the disadvantage of acting as a black box\, not providing\nany informatio
 n about what made them arrive at a particular decision. This work proposes
  a general\nsolution to the problem of understanding classification decisi
 ons by pixel-wise decomposition of non-\nlinear classifiers. We introduce 
 a methodology that allows to visualize the contributions of single pixels\
 nto predictions for kernel-based classifiers over Bag of Words features an
 d for deep neural networks.\nApplications in computer vision and beyond ar
 e provided. \n
LOCATION:Engineering Department\, CBL Room BE-438
END:VEVENT
END:VCALENDAR