BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:A robust and scalable approach to Bayesian doubly-intractable prob
 lems - Francois-Xavier Briol (University College London)
DTSTART:20230309T103000Z
DTEND:20230309T111500Z
UID:TALK196441@talks.cam.ac.uk
DESCRIPTION:Modern Bayesian statistics and machine learning tools are bein
 g applied to increasingly complex physical and biological phenomenon\, and
  as a result make use of increasingly complex models. One such class of mo
 dels are so-called "doubly intractable" models\, for which the likelihood 
 function is known only up to normalisation constant. Examples are as varie
 d as models of multivariate count data arising in genomics\, lattice model
 s arising in statistical physics\, or even large protein signalling networ
 k models arising in biochemistry. Unfortunately\, the Bayesian treatment o
 f such problems presents two main challenges. Firstly\, the size of these 
 models and lack of tractability of the likelihood creates significant comp
 utational challenges\, with standard MCMC or variational methods not direc
 tly applicable. Secondly\, the complexity of the underlying phenomena mean
 s that the models proposed by scientists are often partly incomplete\, and
  as a result misspecified. To solve these issues\, we propose a novel clas
 s of generalised Bayesian posteriors\, which depart from the classical Bay
 esian approach by updating beliefs through loss functions instead of likel
 ihoods. We will show how this approach allows us to select loss functions 
 which provide both computational tractability and robustness to misspecifi
 cation\, and illustrate the approach on examples in genomics\, physics and
  biochemistry which are beyond the scope of current techniques.
LOCATION:Seminar Room 2\, Newton Institute
END:VEVENT
END:VCALENDAR