BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Approximate the Simulator\, Not the Inference: Using Converging Ha
 miltonian Monte Carlo for Flexible SBI in Astronomy - Matthew O'Callaghan 
 (IoA\, Cambridge)
DTSTART:20250210T160000Z
DTEND:20250210T170000Z
UID:TALK223066@talks.cam.ac.uk
CONTACT:65128
DESCRIPTION:Reliably including computationally intensive simulators with i
 ntractable densities in an inference problem is of paramount importance in
  astronomy. In this talk\, I present a normalizing flow architecture that 
 can replace intractable simulators in Bayesian models\, while guaranteeing
  fast convergence of the No U-Turn Sampler variant of Hamiltonian Monte Ca
 rlo (HMC) to the posterior distribution. I discuss the limitations of usin
 g flow-based methods in HMC\, and illustrate examples of a correct impleme
 ntation for toy examples. I conclude by illustrating its use for estimatin
 g interstellar extinction from stellar colours.
LOCATION:Martin Ryle Seminar Room\, KICC
END:VEVENT
END:VCALENDAR