BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:b51d2830c2f2acc7eda835dad41f3ae3
DTSTAMP:20260405T171334Z
SUMMARY:Ben Calderhead – A general construction for parallelising Metropo
 lis-Hastings algorithms
DESCRIPTION:A general construction for parallelising Metropolis-Hastings al
 gorithms: with application to differential equation modelling of biochemi
 cal systems\n Markov chain Monte Carlo methods are essential tools for sol
 ving many modern day statistical and computational problems\, however a m
 ajor limitation is the inherently sequential nature of these algorithms.
  In this talk we propose a natural generalisation of the Metropolis-Hast
 ings algorithm that allows for parallelising a single chain using existin
 g MCMC samplers\, while maintaining convergence to the correct stationary
  distribution. We do so by proposing multiple points in parallel\, then
  constructing and sampling from a finite state Markov chain on the propos
 ed points that has the correct target density as its stationary distribut
 ion. Our approach is generally applicable and easy to implement. We dem
 onstrate how this construction may be used to greatly increase the comput
 ational speed of a wide variety of existing MCMC methods\, including Metr
 opolis-Adjusted Langevin Algorithms\, Adaptive MCMC and Hamiltonian Monte
  Carlo. As a motivating example\, we consider Bayesian inference of biol
 ogical systems described using nonlinear differential equation models.
URL:https://www.imperial.ac.uk/events/106454/ben-calderhead-a-general-const
 ruction-for-parallelising-metropolis-hastings-algorithms/
DTSTART;TZID=Europe/London:20140506T130000
DTEND;TZID=Europe/London:20140506T140000
LOCATION:United Kingdom
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
DTSTART:20140506T130000
TZNAME:BST
TZOFFSETTO:+0100
TZOFFSETFROM:+0100
END:DAYLIGHT
END:VTIMEZONE
END:VCALENDAR