BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Bayesian tangent space inference for stochastic epidemiological mo
 dels - Guenther Turk\, DAMTP
DTSTART:20200512T120000Z
DTEND:20200512T130000Z
UID:TALK142105@talks.cam.ac.uk
CONTACT:Etienne Fodor
DESCRIPTION:The pandemic of the novel coronavirus\, COVID-19\, has hugely 
 affected billions of people worldwide. Forecasting the course of an epidem
 ic based on limited case data\, when viewed as a problem of Bayesian infer
 ence\, is beset by uncertainties in both the epidemiological model and its
  parameters. We assess the effect of these uncertainties for a family of a
 ge-structured compartment models. By formulating these models in terms of 
 a chemical Master equation and taking a diffusion limit\, we readily obtai
 n a quasi-analytical expression for the parameter posterior distribution i
 n the tangent space to the epidemiological solution manifold.  The traceab
 le form of the posterior distribution enables us to compute the Hessian ma
 trix by automatic differentiation.  Therefore\, we can quantify the uncert
 ainty of our parameter estimates exactly\, and obtain a Laplacian approxim
 ation to the Bayesian model evidence.
LOCATION:Zoom: https://zoom.us/j/5916271322 Meeting ID: 591-627-1322
END:VEVENT
END:VCALENDAR