BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Hamiltonian Monte Carlo for Hierarchical Models - Vidhi Lalchand
DTSTART:20191023T130000Z
DTEND:20191023T143000Z
UID:TALK133699@talks.cam.ac.uk
CONTACT:Robert Pinsler
DESCRIPTION:Hierarchical models provide a powerful framework for modelling
  and inference by defining second order and third order probability distri
 butions over parameters at different levels of the generative model. Hamil
 tonian Monte Carlo (HMC) is one of the primary tools for inference in hier
 archical models. While hierarchies provide modelling flexibility\, they in
 duce distinctive pathologies in the posterior that limit the efficiency of
  sampling algorithms like HMC. These pathologies can be best detected by v
 isualising the joint posterior\ngeometry through bivariate density plots a
 nd by HMC diagnostics. In this talk we will review HMC and its limitations
  in the context of posterior inference in hierarchical models. We will dis
 cuss some common techniques to simplify the posterior geometry through\nre
 parameterization that can significantly improve sampling efficiency. We wi
 ll also briefly review advances like Riemann Manifold HMC that can address
  some of the weaknesses of Euclidean HMC in sampling from posterior geomet
 ries characterized by tight correlations and\ndrastically changing curvatu
 re.
LOCATION:Engineering Department\, CBL Room BE-438
END:VEVENT
END:VCALENDAR