BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Models and inference for temporal Gaussian processes - William Wil
 kinson\, Aalto University\, Finland
DTSTART:20191112T110000Z
DTEND:20191112T120000Z
UID:TALK134572@talks.cam.ac.uk
CONTACT:Dr R.E. Turner
DESCRIPTION:Reformulating Gaussian process (GP) models as stochastic diffe
 rential \nequations (SDE) opens up new opportunities to draw comparisons b
 etween classical signal processing algorithms and the state of the art \ni
 nference methods that dominate the current machine learning literature. \n
 In this talk\, I'll discuss the SDE form of the popular "spectral \nmixtur
 e" GP\, showing how in the temporal case we can think of such a \nmodel as
  performing probabilistic time-frequency analysis. Following \nthis\, I wi
 ll discuss recent advances in approximate inference for \ntemporal GP mode
 ls\, focusing on power expectation propagation (EP)\, \nwhich is a natural
  fit for sequential data. Lastly\, I will present new \nwork showing the c
 onditions under which power EP recovers traditional \nmethods such as the 
 extended Kalman filter.
LOCATION:Engineering Department\, CBL Room BE-438.
END:VEVENT
END:VCALENDAR