BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Parallel Markov Chain Monte Carlo - Schmidler\, S (Duke University
 )
DTSTART:20140424T135500Z
DTEND:20140424T143000Z
UID:TALK52215@talks.cam.ac.uk
CONTACT:Chie Sibley Obata
DESCRIPTION:Co-author: Doug VanDerwerken (Duke University) \n\nMarkov chai
 n Monte Carlo is an inherently serial algorithm. Although the likelihood c
 alculations for individual steps can sometimes be parallelized\, the seria
 l evolution of the process is widely viewed as incompatible with paralliza
 tion\, offering no speedup for sampling algorithms which require large num
 bers of iterations to converge to equilibrium. We provide a methodology fo
 r parallelizing Markov chain Monte Carlo across large numbers of independe
 nt\, asynchronous processors. The method is originally motivated by sampli
 ng multimodal target distributions\, where we see an exponential speed-up 
 in running time. However we show that the approach is general purpose and 
 applicable to all Markov chain Monte Carlo simulations\, and demonstrate s
 peed-ups proportional to the number of available processors on slowly mixi
 ng chains with unimodal target distributions. The approach is simple and e
 asy to implement\, and suggests additional directions for further research
 .
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR