BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Rectifying Conformity Scores for Better Conditional Coverage - Max
 im Panov (Mohamed bin Zayed University of Artificial Intelligence)
DTSTART:20250604T110500Z
DTEND:20250604T112500Z
UID:TALK230809@talks.cam.ac.uk
DESCRIPTION:We present a new method for generating confidence sets within 
 the split conformal prediction framework. Our method performs a trainable 
 transformation of any given conformity score to improve conditional covera
 ge while ensuring exact marginal coverage. The transformation is based on 
 an estimate of the conditional quantile of conformity scores. The resultin
 g method is particularly beneficial for constructing adaptive confidence s
 ets in multi-output problems where standard conformal quantile regression 
 approaches have limited applicability. We develop a theoretical bound that
  captures the influence of the accuracy of the quantile estimate on the ap
 proximate conditional validity\, unlike classical bounds for conformal pre
 diction methods that only offer marginal coverage. We experimentally show 
 that our method is highly adaptive to the local data structure and outperf
 orms existing methods in terms of conditional coverage\, improving the rel
 iability of statistical inference in various applications.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR