BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Scalable Monte Carlo inference for state-space models - Sinan Yild
 irim (Sabanci University)
DTSTART:20170706T143000Z
DTEND:20170706T151500Z
UID:TALK73178@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:<span>Co-authors: Christophe Andrieu		(University of Bristol)\
 , Arnaud Doucet		(University of Oxford)        <br></span><span><br>We pre
 sent an original simulation-based method to estimate likelihood ratios eff
 iciently for general state-space models. Our method relies on a novel use 
 of the conditional Sequential Monte Carlo (cSMC) algorithm introduced in A
 ndrieu et al. (2010) and presents several practical advantages over standa
 rd approaches. The ratio is estimated using a unique source of randomness 
 instead of estimating separately the two likelihood terms involved. Beyond
  the benefits in terms of variance reduction one may expect in general fro
 m this type of approach\, an important point here is that the variance of 
 this estimator decreases as the distance between the likelihood parameters
  decreases. We show how this can be exploited in the context of Monte Carl
 o Markov chain (MCMC) algorithms\, leading to the development of a new cla
 ss of exact-approximate MCMC methods to perform Bayesian static parameter 
 inference in state-space models. We show through simulations that\, in con
 trast to the Particle Mar ginal Metropolis&ndash\;Hastings (PMMH) algorith
 m of Andrieu et al. (2010)\, the computational effort required by this nov
 el MCMC scheme scales favourably for large data sets.</span>
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR