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:Particle islands and archipelagos: some large samp
le theory - Olsson\, JR (KTH - Royal Institute of
Technology)
DTSTART;TZID=Europe/London:20140424T103000
DTEND;TZID=Europe/London:20140424T110500
UID:TALK52164AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/52164
DESCRIPTION:Co-authors: Christelle VergĂ©\, Pierre Del Moral\,
and Eric Moulines \n\nThis talk discusses parallel
isation of sequential Monte Carlo methods via the
particle island framework (VergĂ© et al.\, 2013) an
d presents some novel convergence results for meth
ods of this sort. More specifically\, we introduce
the concept of weighted archipelagos (i.e. sets o
f weighted particle islands\, where each island is
itself a weighted sample of particles) and define
three different operations on such archipelagos\,
namely: selection on the island level\, selection
on the particle level\, and mutation. We then est
ablish that these operations preserve a set of con
vergence properties\, including asymptotic normali
ty\, of the archipelago as the number of islands a
s well as the number of particles of each island t
end jointly to infinity. Moreover\, we provide rec
ursive formulas for the asymptotic variance associ
ated with each operation. As our results allow arb
itrary compositions of the mentioned operations to
be analysed\, we may use the same for establishin
g the convergence properties of not only the doubl
e bootstrap algorithm but also generalisations of
this algorithm.
LOCATION:Seminar Room 1\, Newton Institute
CONTACT:Mustapha Amrani
END:VEVENT
END:VCALENDAR