BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Laplace approximation and robust Monte Carlo for Bayesian inferenc
 e with infomative data
DTSTART:20210520T161500
DTEND:20210520T171500
DTSTAMP:20260408T034948Z
UID:5d5ea7cd3674d986103290f781f2f5022e64fdb1ef3be66976136afb
CATEGORIES:Conferences - Seminars
DESCRIPTION:Bjoern Sprungk (TU Bergakademie Freiberg)\nIn Bayesian statist
 ics and Bayesian inverse problems sampling from or integrating with respec
 t to the resulting posterior distribution can become computationally quite
  challenging\, for instance\, if the parameter space is high-dimensional o
 r if the posterior is highly concentrated. Regarding high-dimensional Baye
 sian inference a lot of effort has been spent on deriving dimension-indepe
 ndent sampling methods in recent years. However\, the challenge of a conce
 ntrated posterior resulting from large or informative data has drawn less 
 attention so far -- despite its importance for practical purposes. In this
  talk\, we exploit the well-known Laplace approximation as a suitable Gaus
 sian reference measure for importance sampling as well as for Markov chain
  Monte Carlo sampling of the posterior. We analyse the statistical efficie
 ncy of these methods for a decaying observational noise in the data\, i.e.
 \, for an increasing concentration of the posterior measure. Besides conve
 rgence in Hellinger distance of the Laplace approximation to the posterior
  (a result closely related to the Bernstein-von Mises theorem) we also sho
 w that the proposed Laplace-based Monte Carlo methods perform robustly w.r
 .t. increasing concentration of the posterior whereas prior-based sampling
  methods do not.\n\nThis is joint work together with Claudia Schillings\, 
 Daniel Rudolf\, and Philipp Wacker.
LOCATION:https://epfl.zoom.us/j/84030108577?pwd=bHh2Z3J2YllvTWdteHA3MHhVcn
 IyUT09
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR