BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Generalization and Informativeness of Conformal Prediction - Dr Fr
 edrik Hellstrom\, UCL
DTSTART:20250205T140000Z
DTEND:20250205T150000Z
UID:TALK223645@talks.cam.ac.uk
CONTACT:Prof. Ramji Venkataramanan
DESCRIPTION:A popular technique for uncertainty quantification is conforma
 l prediction\, which converts point predictions into set predictions that 
 are guaranteed to contain the true label of a test input with a user-defin
 ed probability. However\, the size of the predicted set---and thus the inf
 ormativeness of the prediction---is not controlled. In this talk\, we pres
 ent a theoretical connection between the informativeness of conformal pred
 iction sets and generalization properties of the underlying model. Further
 more\, we extend this analysis to conformal risk control and covariate shi
 fts. The results provide insight into the effect of task-specific quantiti
 es and algorithmic hyperparameters\, which we also illustrate via experime
 nts.
LOCATION:MR5\, CMS Pavilion A
END:VEVENT
END:VCALENDAR