BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Rethinking aleatoric and epistemic uncertainty - Freddie Bickford 
 Smith (University of Oxford)
DTSTART:20260309T160000Z
DTEND:20260309T170000Z
UID:TALK245377@talks.cam.ac.uk
CONTACT:Xianda Sun
DESCRIPTION:The ideas of aleatoric and epistemic uncertainty are widely us
 ed to reason about the probabilistic predictions of machine-learning model
 s. We identify incoherence in existing discussions of these ideas and sugg
 est this stems from the aleatoric-epistemic view being insufficiently expr
 essive to capture all the distinct quantities that researchers are interes
 ted in. To address this we present a decision-theoretic perspective that r
 elates rigorous notions of uncertainty\, predictive performance and statis
 tical dispersion in data. This serves to support clearer thinking as the f
 ield moves forward. Additionally we provide insights into popular informat
 ion-theoretic quantities\, showing they can be poor estimators of what the
 y are often purported to measure\, while also explaining how they can stil
 l be useful in guiding data acquisition.\n\n\n\nPaper: https://proceedings
 .mlr.press/v267/bickford-smith25a.html
LOCATION:Cambridge University Engineering Department\, CBL Seminar room BE
 4-38.
END:VEVENT
END:VCALENDAR