Niwanthi Ruwan Kumari Muthukuda Arachchilage
Title: Annealed Sequential Monte Carlo with Adaptive Multiple-Try Metropolis Kernel and Applications to Disease Transmission Models
Date: Monday, August 12th, 2024
Time: 10:30am
Location: ASB 10920
Supervised by: Dr. Liangliang Wang
Abstract: Analyzing infectious diseases such as COVID-19 and their potential to cause pandemics is of utmost importance in epidemiology. Mathematical and statistical epidemiology utilizes various transmission dynamic models to study the spread of infectious diseases, and these models are often represented as a system of Ordinary Differential Equations (ODEs). Markov Chain Monte Carlo (MCMC) and Sequential Monte Carlo (SMC) can be used for conducting Bayesian inference for the unknown parameters in ODEs. But the standard MCMC and SMC methods become inefficient for complex models with high-dimensional data. To address the computational challenges in transmission models, such as the Susceptible-Infectious-Recovered (SIR) model and the Susceptible-Exposed-Infectious-Recovered (SEIR) model, we propose to integrate an adaptive Multiple-Try Metropolis (MTM) kernel as a proposal between intermediate distributions in the Annealed Sequential Monte Carlo (ASMC) algorithm. A simulation study was performed, implementing the SIR and SEIR models using adaptive MCMC, ASMC, and the proposed algorithm. The results demonstrated the efficiency of our methodology, which achieved comparable results with significantly fewer resources. Moreover, real-data analysis was conducted using the new algorithm on the COVID-19 data from Sri Lanka and British Columbia (BC), Canada, during the first half of 2020.
Keywords: Adaptive Multiple-Try Metropolis, Annealed Sequential Monte Carlo, Infectious Disease Modeling