Multivariate Lattice State-Space Modelling for Environmental Applications 

Rodeschini Jacopo ( University of Bergamo)

Wednesday 19th February 11:00-12:00 Maths 311B

Abstract

We present a novel discrete time-invariant State Space Model (SSM) in which the latent space is a discretely indexed Gaussian Markov random field (GMRF), enabling sparse matrix computation. The approach combines the efficiency of GMRF-based methods with the flexibility of state-space modelling. It supports missing data handling and an efficient EM algorithm. The paper also includes theoretical and numerical analyses of the GMRF approximation accuracy and convergence properties. 

Add to your calendar

Download event information as iCalendar file (only this event)