BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Modulated Bayesian Optimisation - Carl Henrik Ek
DTSTART:20201103T131500Z
DTEND:20201103T141500Z
UID:TALK152638@talks.cam.ac.uk
CONTACT:Mateja Jamnik
DESCRIPTION:"Join us on Zoom":https://zoom.us/j/99166955895?pwd=SzI0M3pMVE
 kvNmw3Q0dqNDVRalZvdz09\n\nProbabilistic nummerics aims at providing a  pri
 ncipled framework of uncertainty to nummerical  methods. Importantly this 
 allows different methods to be  contrasted as their implicit assumptions h
 ave to be made  explicit. In this talk I will introduce the idea of  proba
 bilistic nummerics and provide a formulation of Bayesian  optimisation wit
 hin this framework. In specific I will present a  set of surrogage models 
 that focuses on the details that are  informative for search while ignorin
 g detrimental structure that  are challenging to model from few observatio
 ns. We demonstrate  that surrogate models with appropriate noise distribut
 ions can absorb challenging structures in the objective function by  treat
 ing them as irreducible uncertainty. 
LOCATION:Zoom
END:VEVENT
END:VCALENDAR