BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Diffusions limits of the Random Walk Metropolis Algorithm in High 
 Dimensions: -  Natesh Pillai\, Warwick
DTSTART:20090213T160000Z
DTEND:20090213T170000Z
UID:TALK16635@talks.cam.ac.uk
CONTACT:8419
DESCRIPTION:Metropolis-Hastings methods  form a widely used\nclass of MCMC
  methods for sampling from complex\nprobability distributions. Diffusion l
 imits of MCMC methods\n (obtained by a invariance principle argument)\n in
  high dimensions provide\na useful theoretical tool for studying efficienc
 y.\nIn particular they facilitate precise estimates\nof the number of step
 s required to explore the target measure\,\nin stationarity\, as a functio
 n of the dimension of the state\nspace. However\, to date such results hav
 e only been proved\nfor target measures with a product structure\, severel
 y limiting\ntheir applicability to real applications. In this talk\,\nwe w
 ill discuss diffusion limits for a class of naturally\noccuring high dimen
 sional measures\, found from the approximation\nof measures on a Hilbert s
 pace which are absolutely continuous\nwith respect to a Gaussian reference
  measure.
LOCATION:MR12\, CMS\, Wilberforce Road\, Cambridge\, CB3 0WB
END:VEVENT
END:VCALENDAR