BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Continuous-time Importance Sampling for Multivariate Diffusions - 
 Paul Fearnhead\, Lancaster University
DTSTART:20120608T150000Z
DTEND:20120608T160000Z
UID:TALK37810@talks.cam.ac.uk
CONTACT:Richard Samworth
DESCRIPTION:Inference for multivariate diffusion processes is challenging 
 due to the\nintractability of the dynamics of the process. Most methods re
 ly on high\nfrequency imputation and discrete-time approximations of the c
 ontinuous-time\nmodel\, leading to biased inference.  Recently\, methods t
 hat are able to\nperform inference for univariate diffusions which avoid t
 ime-discretisation\nerrors have been developed. However these approaches c
 annot be applied to\ngeneral multivariate diffusions.\n\nHere we present a
  novel\, continuous-time Importance Sampling method that\nenables inferenc
 e for\ngeneral continuous-time Markov processes whilst avoiding time-discr
 etisation\nerrors. The method\ncan be derived as a limiting case of a disc
 rete-time sequential importance\nsampler\, and uses ideas from random-weig
 ht particle filters\, retrospective\nsampling and Rao-Blackwellisation.\n\
 nJoint work with Gareth Roberts\, Giorgos Sermaidis and Krys Latuszynski.
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0WB
END:VEVENT
END:VCALENDAR