BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Improving conditional coverage of conformal prediction methods - M
 axim Panov (Mohamed bin Zayed University of Artificial Intelligence)
DTSTART:20250506T130000Z
DTEND:20250506T133000Z
UID:TALK230443@talks.cam.ac.uk
DESCRIPTION:We present and compare two new methods for generating predicti
 on sets within the conformal prediction framework\, each addressing the li
 mitations of traditional approaches by targeting improved conditional cove
 rage. The first method builds upon quantile regression to estimate the con
 ditional quantile of conformity scores\, which are then adjusted to accoun
 t for local data structure. The second method integrates the flexibility o
 f conformal methods with estimates of the conditional distribution label d
 istribution​. By extending the framework of probabilistic conformal pred
 iction\, this approach achieves approximately conditional coverage through
  prediction sets that adapt effectively to the behavior of the predictive 
 distribution\, even under high heteroscedasticity. Non-asymptotic bounds a
 re derived to quantify conditional coverage error for both approaches. Ext
 ensive simulations demonstrate that each method significantly improves ove
 r traditional techniques\, paving the way for more robust and adaptable pr
 ediction set generation across diverse applications.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR