BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Isaac Newton Institute Seminar Series
SUMMARY:Efficient implementation of Markov chain Monte Car
lo when using an unbiased likelihood estimator - P
itt\, M (University of Warwick)
DTSTART;TZID=Europe/London:20140422T134500
DTEND;TZID=Europe/London:20140422T142000
UID:TALK52088AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/52088
DESCRIPTION:When an unbiased estimator of the likelihood is us
ed within an Metropolis-Hastings scheme\, it is ne
cessary to tradeoff the number of samples used to
evaluate the likelihood against the computing time
. Many samples will result in a scheme which has s
imilar properties to the case where the likelihood
is exactly known but will be expensive. Few sampl
es will result in faster estimation but at the exp
ense of slower mixing of the Markov chain. We expl
ore the relationship between the number of samples
and the efficiency of the resulting Metropolis-Ha
stings estimates. Under the assumption that the di
stribution of the additive noise introduced by the
log-likelihood estimator is independent of the po
int at which this log-likelihood is evaluated and
other relatively mild assumptions\, we provide gui
delines on the number of samples to select for a g
eneral Metropolis-Hastings proposal. We illustrate
on a complex stochastic volatility model that the
se assumptions are approximately satisfied experim
entally and that the theoretical insights with reg
ards to inefficiency and computational time hold t
rue. \n\nKeywords: Bayesian inference\; Estimated
likelihood\; Metropolis-Hastings\; Particle filter
ing.\n
LOCATION:Seminar Room 1\, Newton Institute
CONTACT:Mustapha Amrani
END:VEVENT
END:VCALENDAR