BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:BSU Seminar: &quot\;Robust and conjugate Gaussian processes&quot\;
  - François-Xavier Briol\, University College London
DTSTART:20251111T140000Z
DTEND:20251111T150000Z
UID:TALK240280@talks.cam.ac.uk
CONTACT:Alison Quenault
DESCRIPTION:To enable closed form conditioning\, a common assumption in Ga
 ussian process (GP) regression is independent and identically distributed 
 Gaussian observation noise.\n This strong and simplistic assumption is oft
 en violated in practice\, which leads to unreliable inferences and uncerta
 inty quantification. Unfortunately\, existing methods for robust GP regres
 sion break closed-form conditioning\, which makes them less attractive\n t
 o practitioners and significantly more computationally expensive. In this 
 talk\, I will discuss a recent line of work which has led to provably robu
 st and conjugate Gaussian process (RCGP) regression at virtually no additi
 onal computational cost using generalised\n Bayesian inference. I will ill
 ustrate the desirable properties of RCGPs for problems ranging from Bayesi
 an optimisation to sparse variational Gaussian process regression\, spatio
 -temporal modelling\, and multi-output regression. The latter will be illu
 strated\n in the context of cancer research\, where we use RCGPs to perfor
 m robust modelling of the viability of cancer cells to varying doses of dr
 ugs.
LOCATION:Large Downstairs Teaching Room\, East Forvie Building\, Forvie Si
 te Robinson Way Cambridge CB2 0SR.
END:VEVENT
END:VCALENDAR