BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Dimension-Robust Function Space MCMC With Neural Network Priors - 
 Torben Sell (University of Cambridge)
DTSTART:20210414T130000Z
DTEND:20210414T140000Z
UID:TALK158176@talks.cam.ac.uk
CONTACT:Neil Deo
DESCRIPTION:At the beginning of this talk\, two popular priors defined on 
 function spaces are discussed: Gaussian priors\, which come with a set of 
 orthogonal basis functions\, and Bayesian Neural Networks (BNNs)\, which a
 re popular in the machine learning community. I argue that both priors com
 e with disadvantages\, and propose a new class of BNN priors that alleviat
 e them. The resulting posteriors are amenable to sampling using Hilbert sp
 ace Markov chain Monte Carlo methods (unlike standard BNNs)\, and scale mo
 re favourably in the dimension of the function’s domain (unlike most Gau
 ssian measures). Some theoretical results as well as numerical illustratio
 ns are presented\, and my talk will end by posing future research directio
 ns. This talk is loosely based on the following preprint: https://arxiv.or
 g/abs/2012.10943.
LOCATION:https://maths-cam-ac-uk.zoom.us/j/95531783868?pwd=U3pPbmYxTXZYRVZ
 MWFBVTkVnWmUvZz09
END:VEVENT
END:VCALENDAR