BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:On the use of non-local priors for joint high-dimensional estimati
 on and selection  - David Rossel\, University of Warwick
DTSTART:20150116T160000Z
DTEND:20150116T170000Z
UID:TALK56938@talks.cam.ac.uk
CONTACT:20082
DESCRIPTION:A main challenge in modern statistics is to devise strategies 
 that are effective in high dimensions. Ideally one would like solutions wh
 ich are on one hand parsimonious\, i.e. help interpret the process that ge
 nerated the data in an easy manner\, but that at the same time yield accur
 ate predictions. As has been well-documented in the literature there is a 
 tension between these two competing goals\, e.g. simpler explanatory model
 s tend to result in higher prediction errors\,\nand accurate predictive mo
 dels tend to be more complex than one would ideally wish for.\n\nWe explor
 e the extent to which these two goals can be reconciled\, adopting a Bayes
 ian framework and a novel formulation based on non-local priors (NLPs).\nT
 his class of priors has been proven to lead to faster learning rates for m
 odel selection that are indispensable if one is to attain Bayesian consist
 ency in high-dimensions.\nBecause they induce extra parsimony in the solut
 ion\, NLPs typically result in adequately simple explanatory models.\nInte
 restingly\, we recently discovered that NLPs also result in improved shrin
 kage rates for parameter estimation that lead to highly accurate predictio
 ns in high dimensions\, hence providing models that are at the same time p
 romising for explanatory and predictive purposes.\nWe will illustrate thes
 e issues by reviewing some of the relevant theory and show various practic
 al examples\,\nas well as propose strategies to deal efficiently with some
  of the main computational issues at stake.
LOCATION:MR12\,  Centre for Mathematical Sciences\, Wilberforce Road\, Cam
 bridge
END:VEVENT
END:VCALENDAR