BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Hessian-based Markov-Chain Monte Carlo Algorithms - Tom Minka (Mic
 rosoft Research Ltd)
DTSTART:20090702T130000Z
DTEND:20090702T143000Z
UID:TALK18078@talks.cam.ac.uk
CONTACT:Shakir Mohamed
DESCRIPTION:Hessian-based Markov-Chain Monte Carlo Algorithms \nTom Minka 
 Microsoft Research Cambridge\n----\nI will talk about how to make Markov-c
 hain Monte Carlo run more efficiently in high-dimensional\, continuous spa
 ces.  The idea is to shape the Markov transition density according to the 
 local Hessian of the probability density function.  This leads to a Hessia
 n-based Metropolis-Hastings algorithm that we call HMH.  A naive implement
 ation of this idea would be quite expensive however\, requiring the Hessia
 n to be recomputed at each sample.  Instead I will describe how to increme
 ntally update the Hessian\, and how to get many samples from the same Hess
 ian (using the multiple-try Metropolis algorithm).  The upshot is that\, g
 iven any function where you can do efficient Hessian-based optimization\, 
 you can also do efficient sampling.\n\nJoint work with Yuan (Alan) Qi.\n\n
 Link to paper:\nhttp://www.cs.purdue.edu/homes/alanqi/papers.html\n
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR