BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Amortized Bayesian experimental design with sequential Monte Carlo
  - Sahel Mohammad Iqbal (Aalto University)
DTSTART:20250624T091500Z
DTEND:20250624T094500Z
UID:TALK232204@talks.cam.ac.uk
DESCRIPTION:Most existing works on amortized Bayesian experimental design 
 (BED) rely on contrastive estimators of the expected information gain (EIG
 ). In this talk\, I introduce a new framework for BED grounded in the cont
 rol-as-inference paradigm. The task of collecting informative trajectories
  is reframed as a sampling problem from a non-Markovian state-space model.
  To address the resulting inference challenges\, I present a nested sequen
 tial Monte Carlo algorithm tailored to this setting. This approach offers 
 a fresh perspective on BED\, and we end with ideas for improving the scala
 bility of this algorithm.
LOCATION:Seminar Room 1\, Newton Institute
END:VEVENT
END:VCALENDAR