BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:&quot\;Design for ABC and history matching with Gaussian processes
 &quot\; - Dr Richard Wilkinson\, Sheffield University
DTSTART:20160301T143000Z
DTEND:20160301T153000Z
UID:TALK63429@talks.cam.ac.uk
CONTACT:Alison Quenault
DESCRIPTION:Abstract: History matching and approximate Bayesian computatio
 n (ABC) are techniques for parameter estimation for complex models. In bot
 h cases\, Gaussian processes can be used to accelerate inference by buildi
 ng an emulator/surrogate model either of the simulator itself\, or of a su
 mmary such as the likelihood function.  Design approaches for building emu
 lators usually focus on how to build accurate global approximations. Howev
 er\, if our aim is calibration\, then this can be wasteful. In this talk I
  will discuss the use of Gaussian processes\, and describe design strategi
 es that can be used for history matching and calibration\, both in a deter
 ministic and stochastic setting. Entropy based designs that minimize the u
 ncertainty in the classification surface are particularly useful\, and I w
 ill show how these can be efficiently computed using Bayesian optimisation
 .
LOCATION:Large  Seminar Room\, 1st Floor\, Institute of Public Health\, Un
 iversity Forvie Site\, Robinson Way\, Cambridge
END:VEVENT
END:VCALENDAR