BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Asymptotics of Approximate Bayesian Computation - Paul Fearnhead (
 Lancaster University)
DTSTART:20170705T080000Z
DTEND:20170705T084500Z
UID:TALK73152@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:Many statistical applications involve models for which are eas
 y to sample from\, but for which it is&nbsp\;difficult to evaluate the lik
 elihood. Approximate Bayesian computation is a likelihood-free method for 
 implementing Bayesian inference in such cases. This talk will overview som
 e recent results on the theoretical properties of approximate Bayesian com
 putation which consider the performance of ABC as we get more data. It wil
 l cover questions such as: when does the ABC posterior concentrate on the 
 true parameter value? What distribution does the ABC posterior converge to
 ? And what is the frequentist distribution of point-estimates derived usin
 g ABC. It will also cover the impact of Monte Carlo error on estimates obt
 ained using ABC\, and consider whether\, asympotically\,&nbsp\;it is possi
 ble to efficiently estimate parameters using ABC if we have a fixed Monte 
 Carlo sample size.<br><br>This is joint work with Wentao Li:&nbsp\;<a targ
 et="_blank" rel="nofollow" href="https://arxiv.org/abs/1506.03481">https:/
 /arxiv.org/abs/1506.03481</a> and&nbsp\;&nbsp\;<a target="_blank" rel="nof
 ollow" href="https://arxiv.org/abs/1609.07135">https://arxiv.org/abs/1609.
 07135</a>\; the talk will also cover work by David Frazier\, Martin\, Robe
 rt and Rousseau:&nbsp\;<a target="_blank" rel="nofollow" href="https://arx
 iv.org/abs/1607.06903">https://arxiv.org/abs/1607.06903</a>
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR