THE UNIVERSITY OF WESTERN AUSTRALIA
Faculty of Engineering, Computing and Mathematics
School of Mathematics and Statistics
Title: Particle Markov chain Monte Carlo methods in marine biogeochemistry
Presenter: Lawrence Murray
CSIRO Mathematics, Informatics & Statistics
Leeuwin Centre, Floreat, WA
Date: Thursday 13 September 2012
Time:2.00 – 3:00 pm
Location: Mathematics Lecture Room 3 (G.02, Ground floor of Mathematics Building (entrance from the James oval side))
I'll report on progress in applying particle Markov chain Monte Carlo (PMCMC) methods for state and parameter estimation in the domain of marine biogeochemistry, motivated by a desire to quantify uncertainty in forecasts used for environmental management. Process models in this area combine physical ocean models with highly nonlinear biology, and may have significant dimensionality. Such models can be coupled to data using a state-space model framework, and Bayesian inference performed using PMCMC.
I'll highlight advances in dealing with some specific challenges that these cases pose: overcoming the absence of a closed-form transition density and ameliorating computational expense on modern GPU hardware. Along the way we draw out some general lessons as to the behaviour of PMCMC methods in the wild, and strategies for improving their performance on this difficult class of problems.
All are welcome to attend the seminar
No RSVP required.
Enquiries: Gopalan Nair (08) 6488 3377; [email protected]