BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Forward Smoothing using Sequential Monte Carlo with Application to
     Recursive Parameter Estimation - Arnaud Doucet  (Univ. of British Colu
 mbia and ISM Tokyo)
DTSTART:20091020T140000Z
DTEND:20091020T150000Z
UID:TALK20463@talks.cam.ac.uk
CONTACT:Richard Nickl
DESCRIPTION:Sequential Monte Carlo (SMC) methods are a widely used set of\
 ncomputational tools for inference in non-linear non-Gaussian state-space\
 nmodels. We propose a new SMC algorithm to compute the expectation of\nadd
 itive functionals recursively. Compared to the standard path space SMC\nes
 timator whose asymptotic variance increases quadratically with time even\n
 under favourable mixing assumptions\, the asymptotic variance of the propo
 sed\nSMC estimator only increases linearly with time. We show how this all
 ows us\nto perform recursive parameter estimation using SMC algorithms whi
 ch do not\nnot suffer from the particle path degeneracy problem.\n\nJoint 
 work with P. Del Moral (INRIA Bordeaux) & S.S. Singh (Cambridge\nUniversit
 y)\n
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0WB
END:VEVENT
END:VCALENDAR