BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Stein Points: Efficient sampling from posterior distributions by m
 inimising Stein Discrepancies. - Francois-Xavier Briol (Imperial College L
 ondon\; University of Warwick\; University of Oxford)
DTSTART:20180523T100000Z
DTEND:20180523T120000Z
UID:TALK107017@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:An important task in computational statistics and machine lear
 ning is to approximate a posterior distribution with an empirical measure 
 supported on a set of representative points. This work focuses on methods 
 where the selection of points is essentially deterministic\, with an empha
 sis on achieving accurate approximation when the number of points is small
 . To this end\, we present `Stein Points&#39\;. The idea is to exploit eit
 her a greedy or a conditional gradient method to iteratively minimise a ke
 rnel Stein discrepancy between the empirical measure and the target measur
 e. Our empirical results demonstrate that Stein Points enable accurate app
 roximation of the posterior at modest computational cost. In addition\, th
 eoretical results are provided to establish convergence of the method.
LOCATION:Seminar Room 2\, Newton Institute
END:VEVENT
END:VCALENDAR