BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Robust inference with the knockoff filter - Rina Foygel Barber (Un
 iversity of Chicago)
DTSTART:20180115T111000Z
DTEND:20180115T115500Z
UID:TALK97567@talks.cam.ac.uk
CONTACT:INI IT
DESCRIPTION:In this talk\, I will present ongoing work on the knockoff fil
 ter for inference in regression. In a high-dimensional model selection pro
 blem\, we would like to select relevant features without too many false po
 sitives. The knockoff filter provides a tool for model selection by creati
 ng knockoff copies of each feature\, testing the model selection algorithm
  for its ability to distinguish true from false covariates to control the 
 false positives. In practice\, the modeling assumptions that underlie the 
 construction of the knockoffs may be violated\, as we cannot know the exac
 t dependence structure between the various features. Our ongoing work aims
  to determine and improve the robustness properties of the knockoff framew
 ork in this setting. We find that when knockoff features are constructed u
 sing estimated feature distributions whose errors are small in a KL diverg
 ence type measure\, the knockoff filter provably controls the false discov
 ery rate at only a slightly higher level. This work is joint with Emmanuel
  Cand&egrave\;s and Richard Samworth.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR