BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Isaac Newton Institute Seminar Series
SUMMARY:A robust and scalable approach to Bayesian doubly-
 intractable problems - Francois-Xavier Briol (Univ
 ersity College London)
DTSTART;TZID=Europe/London:20230309T103000
DTEND;TZID=Europe/London:20230309T111500
UID:TALK196441AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/196441
DESCRIPTION:Modern Bayesian statistics and machine learning to
 ols are being applied to increasingly complex phys
 ical and biological phenomenon\, and as a result m
 ake use of increasingly complex models. One such c
 lass of models are so-called "doubly intractable" 
 models\, for which the likelihood function is know
 n only up to normalisation constant. Examples are 
 as varied as models of multivariate count data ari
 sing in genomics\, lattice models arising in stati
 stical physics\, or even large protein signalling 
 network models arising in biochemistry. Unfortunat
 ely\, the Bayesian treatment of such problems pres
 ents two main challenges. Firstly\, the size of th
 ese models and lack of tractability of the likelih
 ood creates significant computational challenges\,
  with standard MCMC or variational methods not dir
 ectly applicable. Secondly\, the complexity of the
  underlying phenomena means that the models propos
 ed by scientists are often partly incomplete\, and
  as a result misspecified. To solve these issues\,
  we propose a novel class of generalised Bayesian 
 posteriors\, which depart from the classical Bayes
 ian approach by updating beliefs through loss func
 tions instead of likelihoods. We will show how thi
 s approach allows us to select loss functions whic
 h provide both computational tractability and robu
 stness to misspecification\, and illustrate the ap
 proach on examples in genomics\, physics and bioch
 emistry which are beyond the scope of current tech
 niques.
LOCATION:Seminar Room 2\, Newton Institute
CONTACT:
END:VEVENT
END:VCALENDAR