BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Introduction to Gaussian processes and Modulated Bayesian Optmisat
 ion - Carl Henrik Ek 	
DTSTART:20200803T153000Z
DTEND:20200803T163000Z
UID:TALK149944@talks.cam.ac.uk
CONTACT:Bingqing Cheng
DESCRIPTION:*** Abstract\n\nThis talk will be divided into two parts. In t
 he first part I will\nprovide an introduction to stochastic processes and 
 in specific\nGaussian processes. We will see how these mathematically simp
 le\nobjects can be used to specify distributions over the space of\nfuncti
 ons. This allows for Bayesian inference over function\nproviding a princip
 led treatment of uncertainty.\n\nThe second part of the talk will focus on
  sequential decision\nmaking under uncertainty. In specific we will discus
 s Bayesian\nOptmisation (BO) where the goal is to find the extremum of an\
 nexplicitly unknown function. By placing a GP-prior as a surrogate\nto the
  unknown function we can formulate optimisation as a search\nstrategy. How
 ever\, the BO methodology suffers from the fact that\nthe modelling task i
 s distinct from the search. This often leads\nto sub-optimal decisions whe
 n structures irrelevant for the optima\nhave been observed. In this talk I
  will describe recent worked\ncalled "Modulating Surrogates for Bayesian O
 ptmisation". The goal\nof this paper is to provide a surrogate model that 
 allows only\nstructures that are relevant for the search to be modelled.\n
 \n*** bio\n\nDr. Carl Henrik Ek is a recently appointed senior lecturer in
 \nComputer Science at the University of Cambridge. Prior to\nCambridge he 
 worked at the University of Bristol. His research\nfocuses on Bayesian non
 -parametrics and stochastic processes. In\nspecific he is interested in bu
 ilding models for scenarios where\ndata is scares and uncertain for critic
 al applications that\nrequires statistical models that can be interrogated
  and\ninterpret-able.
LOCATION:virtual ZOOM meeting ID: 263 591 6003\, https://zoom.us/j/2635916
 003
END:VEVENT
END:VCALENDAR
