BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Learning a regularisation functional - Sebastian Lunz (University 
 of Cambridge)
DTSTART:20180207T160000Z
DTEND:20180207T170000Z
UID:TALK101173@talks.cam.ac.uk
CONTACT:Ollie McEnteggart
DESCRIPTION:In the light of the immense success of artificial neural netwo
 rks in machine learning computer vision\, hopes have been raised that thes
 e methods applied to inverse problems could also lead to better performanc
 e. We discuss the limitations of existing algorithms for deep learning in 
 inverse problems and then consider the approach of training a neural netwo
 rk as a regularisation functional. We will propose a new algorithm that re
 lies on recent advances in generative modelling using adversarial networks
 . We analyse theoretical properties of this algorithm and present first co
 mputational results.
LOCATION:MR14\, Centre for Mathematical Sciences
END:VEVENT
END:VCALENDAR