BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:BSU Seminar: “Cox-process representation and inference for stoch
 astic reaction-diffusion processes” - Dr Guido Sanguinetti\, University 
 of Edinburgh
DTSTART:20180510T133000Z
DTEND:20180510T143000Z
UID:TALK102913@talks.cam.ac.uk
CONTACT:Alison Quenault
DESCRIPTION:Complex behaviour in many systems arises from the stochastic i
 nteractions of spatially distributed particles or agents. Stochastic react
 ion-diffusion processes are widely used to model such behaviour in discipl
 ines ranging from biology to the social sciences\, yet they are notoriousl
 y difficult to simulate and calibrate to observational data. On the other 
 hand\, spatio-temporal point processes offer several computational advanta
 ges from the statistical perspective\, and can be coupled to dynamics via 
 an evolving intensity field. In this talk\, I will discuss how joint time 
 marginals of a stochastic reaction-diffusion process can be approximated i
 n a mean-field sense by a spatio-temporal Cox process. The resulting appro
 ximation allows us to naturally define an approximate likelihood\, which c
 an be optimised with respect to the kinetic parameters of the model. We sh
 ow on several examples from systems biology and epidemiology that the meth
 od yields consistently accurate parameter estimates\, and can be used effe
 ctively for model selection. \nRef: Schnoerr\, Grima and Sanguinetti\, Nat
 ure Communications 7\, 2016 also\nhttp://arxiv.org/abs/1601.01972\n
LOCATION:Large Seminar Room\, 1st Floor\, Institute of Public Health\, Uni
 versity Forvie Site\, Robinson Way\, Cambridge
END:VEVENT
END:VCALENDAR