BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Speeding up MCMC by Efficient Data Subsampling - Kohn\, R (Univers
 ity of New South Wales)
DTSTART:20140423T093000Z
DTEND:20140423T100500Z
UID:TALK52125@talks.cam.ac.uk
CONTACT:Mustapha Amrani
DESCRIPTION:Co-authors: Chris carter (University of New South wales )\, Ed
 uardo Mendes (University of New South wales ) \n\nThe computing time for M
 arkov Chain Monte Carlo (MCMC) algorithms can be prohibitively large for d
 atasets with many observations\, especially when the data density for each
  observation is costly to evaluate. We propose a framework based on a Pseu
 do-marginal MCMC where the likelihood function is unbiasedly estimated fro
 m a random subset of the data\, resulting in substantially fewer density e
 valuations. The subsets are selected using efficient sampling schemes\, su
 ch as Probability Proportional-to-Size (PPS) sampling where the inclusion 
 probability of an observation is proportional to an approximation of its c
 ontribution to the likelihood function. We illustrate the method on a larg
 e dataset of Swedish firms containing half a million observations.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR