时 间：2025年5月13日（周二）14：30 – 15：30
地 点：理科大楼A614室
报告人：王亮亮 Simon Fraser University（加拿大）Associate Professor
主持人：吴贤毅 华东师范大学教授
摘 要：
Parameter Estimation in Ordinary Differen- tial Equation Models via Sequential Monte Carlo Methods: Ordinary Differential Equations (ODEs) are fundamental in modeling dynamic processes across numerous disciplines, including epidemiology, biology, physics, and economics. Yet, parameter estimation and inference in ODE-based models remain challenging due to inherent nonlinearities, high-dimensional parameter spaces, multimodal likelihood surfaces, and noisy observational data. This talk introduces two of our proposed computational methods to address these chal- lenges. First, we present an adaptive Annealed Sequential Monte Carlo (ASMC) algorithm enhanced by an adaptive Multiple-Try Metropolis (MTM) kernel, specifically developed for complex epidemiological transmission models such as Susceptible-Infectious-Recovered (SIR) and Susceptible-Exposed-Infectious- Recovered (SEIR). The proposed method significantly improves computational efficiency and posterior exploration compared to conventional MCMC and stan- dard SMC algorithms. We demonstrate its effectiveness through simulation studies and real-world analyses of COVID-19 data from British Columbia and Sri Lanka. Second, we introduce Particle Data Cloning (PDC), an innovative method combining data cloning techniques with ASMC to mitigate multimodal- ity in likelihood surfaces, achieving superior global optimization and stable frequentist inference. Simulation studies and practical applications, including predator-prey dynamic systems, illustrate PDC’s reliability and efficiency.
报告人简介：
王亮亮, 加拿大Simon Fraser University统计与精算学系副教授, 研究领域包括Bayesian statistics; Sequential Monte Carlo; Machine learning; Dynamical models; Functional data analysis; Phylogenetics; Biostatistics.