BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Gaussian processes for inferring latent functions in complex data 
 models - Martin Tegner
DTSTART:20181218T103000Z
DTEND:20181218T110000Z
UID:TALK116275@talks.cam.ac.uk
CONTACT:Dr R.E. Turner
DESCRIPTION:Heavily influenced by the challenges in local volatility model
 ling from quantitative finance\, we consider inferring a latent function i
 n a probabilistic model of data. In vicinity of the former\, estimation is
  approached by deterministic means\, commonly least-squares optimisation. 
 Our contribution is to introduce a probabilistic framework based on Gaussi
 an process priors. We approach inference with Markov chain Monte Carlo and
  extend some of these techniques to scale with data. We propose an approxi
 mation that enables sequential sampling of both latent variables and assoc
 iated hyperparameters. We demonstrate our approach through the local volat
 ility model in a growing data settings which would otherwise be unfeasible
  with naive\, non-sequential sampling.
LOCATION:Engineering Department\, CBL Room BE-438.
END:VEVENT
END:VCALENDAR