BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Optimal Monte Carlo Sampling and a Multi-Core Metropolis-Hastings 
 Algorithm - Sam Power 
DTSTART:20191030T160000Z
DTEND:20191030T170000Z
UID:TALK134233@talks.cam.ac.uk
CONTACT:Jan Bohr
DESCRIPTION:*Abstract.* When tasked with drawing approximate samples from 
 a distribution of interest\, perhaps the most commonly-used approach is th
 e Metropolis-Hastings algorithm\, a Markov Chain Monte Carlo (MCMC)techniq
 ue. In this method\, one proceeds by i) proposing moves to new locations\,
  and then ii) deciding whether or not to accept these proposals. A key tra
 deoff with such algorithms concerns how ambitious to be with these proposa
 l moves\; one prefers to make large moves when possible\, but one also wan
 ts to have a reasonable proportion of proposal moves accepted.\n\nIn this 
 talk\, I will review a framework which has provided practically useful ans
 wers on how to navigate this tradeoff\, the theory of `optimal scaling' fo
 r MCMC.I will then discuss recent work on how this theory can be adapted t
 o the scenario in which parallel computing resources are available\, and d
 escribe how one can use them to improve the efficiency of standard MCMC al
 gorithms. This leads to concrete practical recommendations\, as well as pr
 oviding some quantitative estimates for how much benefit one can asymptoti
 cally expect from parallelism.\n
LOCATION:MR14\, Centre for Mathematical Sciences
END:VEVENT
END:VCALENDAR