BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Fast hybrid tempered ensemble transform filter for Bayesian ellipt
 ical problems - Jana de Wiljes (Universität Potsdam)
DTSTART:20230308T150000Z
DTEND:20230308T160000Z
UID:TALK198058@talks.cam.ac.uk
DESCRIPTION:Bayesian inverse problems can be challenging to solve when wor
 king with partial and noisy data\, particularly in high-dimensional and no
 nlinear settings. One commonly used method is ensemble Kalman filtering\, 
 which provides robust and computationally efficient estimations through Ga
 ussian approximations. However\, this method may not accurately approximat
 e non-Gaussian posterior distributions.\nTo address this issue\, the tempe
 red ensemble transform particle filter has been developed as an adaptive s
 equential Monte Carlo method that uses optimal transport mapping for resam
 pling. This approach does not rely on assumptions about the posterior dist
 ribution\, making it suitable for nonlinear non-Gaussian inverse problems.
  However\, it is computationally complex and less robust than ensemble Kal
 man filtering for high-dimensional problems.\nTo improve the accuracy and 
 efficiency of this method\, an entropy-inspired regularization factor to r
 educe computational costs through Sinkhorn iterations is introduced. Addit
 ionally\, we incorporate an ensemble Kalman filtering proposal step before
  each sample update\, resulting in a hybrid approach that further enhances
  the method's robustness.
LOCATION:Seminar Room 2\, Newton Institute
END:VEVENT
END:VCALENDAR