BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Expectation Propagation\, Experimental Design for the Sparse Linea
 r Model - Matthias Seeger (Max Planck Institute for Biological Cybernetics
 )
DTSTART:20080220T140000Z
DTEND:20080220T150000Z
UID:TALK10729@talks.cam.ac.uk
CONTACT:Zoubin Ghahramani
DESCRIPTION:Expectation propagation (EP) is a novel variational method for
  approximate\nBayesian inference\, which has given promising results in te
 rms of computational\nefficiency and accuracy in several machine learning 
 applications. It can readily\nbe applied to inference in linear models wit
 h non-Gaussian priors\, generalised\nlinear models\, or nonparametric Gaus
 sian process models\, among others.\nI will give an introduction to this f
 ramework. Important\naspects of EP are not well-understood theoretically.\
 nI will highlight some open problems.\n\nI will then show how to address s
 equential experimental design for a linear model\nwith non-Gaussian sparsi
 ty priors\, giving some results in two different machine\nlearning applica
 tions. These results indicate that experimental design for these\nmodels m
 ay have significantly different properties than for linear-Gaussian models
 \,\nwhere Bayesian inference is analytically tractable and experimental de
 sign seems best understood.\n
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR