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:Probabilistic Systems\, Information\, and Inferenc
 e Group Seminars
SUMMARY:Optimal proposal densities for particle filters - 
 Peter Jan van Leeuwen\, Prof in Data Assimilation\
 , Department of Meteorology\, University of Readin
 g
DTSTART;TZID=Europe/London:20120125T141500
DTEND;TZID=Europe/London:20120125T150000
UID:TALK35375AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/35375
DESCRIPTION:In this talk\, we discuss why the so-called 'optim
 al proposal density'in particle filtering is not e
 fficient when the number of independent observatio
 ns is large. Two new methods for constructing prop
 osal\ndensities are introduced that ensure that al
 l -or most- of the particles end up in the high pr
 obability regions with similar weights. These simi
 lar weights opens up the field of efficient partic
 le filtering in\nhigh-dimensional systems. Applica
 tions to simple systems and a high-dimensional hig
 hly nonlinear fluid-dynamics problem (the barotrop
 ic vorticity equation) will be discussed.\n\n
LOCATION:LR4\, Engineering\, Department of
CONTACT:Rachel Fogg
END:VEVENT
END:VCALENDAR