BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Keynote: Judging uncertainty from black-box classifiers - Gaël Va
 roquaux (INRIA)
DTSTART:20250827T083000Z
DTEND:20250827T093000Z
UID:TALK234487@talks.cam.ac.uk
DESCRIPTION:A predictive model should ideally express its uncertainty as a
  probability of the output given the input. This is particularly important
  in high-stakes applications such as health. Predictions from black boxes 
 come with many sources of potential uncertainty and error. Some uncertaint
 y arises because the data does not explain perfectly the outcome. Some unc
 ertainty comes from uncertainty on which functional form to use in the pre
 dictive model.I will discuss how analyse uncertainty from black-box classi
 fier. The literature discusses much a quantity know as calibration. Howeve
 r\, full uncertainty requires to go further and control the reminder\, the
  "grouping loss"\, which leads to challenging estimation problems. I'll di
 scuss how to estimate it\, and how it connects to suboptimality gap in a d
 ecision-theory setting.Finally\, I'll do a quick tangent on table foundati
 on models\, which can be seen as a very broad way of designing complex and
  structured priors for predictors.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR