Abstract: Bayesian phylogenetic inference provides a principled framework for reconstructing evolutionary relationships from genomic data, yet the posterior distribution over tree space is highly structured, discrete, and extremely high-dimensional. As a result, standard Markov chain Monte Carlo (MCMC) algorithms often mix poorly or become trapped in local modes, limiting both scalability and inferential accuracy. In this talk, I present a suite of advanced Monte Carlo methodologies designed to overcome these computational bottlenecks. I first introduce our developments in Sequential Monte Carlo (SMC), including an annealed SMC framework and Particle Gibbs strategies specifically adapted to phylogenetic tree spaces. By propagating and adaptively resampling a population of particles, these methods enable efficient exploration of highly multimodal posterior landscapes. I then discuss new computational frameworks for complex posterior sampling, including a unified construction of Multiple-Try Metropolis, the Compound Auxiliary Metropolis algorithm, and Parallel Annealed Arrayed Chains. By incorporating multi-candidate proposals and auxiliary variables, these approaches significantly improve sampling efficiency and robustness in rugged tree spaces. Together, these methods provide scalable and reliable tools for modern Bayesian phylogenetics, ultimately supporting more accurate and computationally feasible genomic inference in evolutionary biology and statistical genetics.
Bio: Liangliang Wang is an Associate Professor in the Department of Statistics and Actuarial Science at Simon Fraser University, where she has been a faculty member since 2013. Dr. Wang earned her Ph.D. in Statistics from the University of British Columbia, following graduate
studies at McGill University and Peking University, and a B.Sc. in Computer Science from Zhengzhou University. Her research advances computational statistics and statistical machine learning, with a specific focus on scalable Bayesian inference, Monte Carlo methods, and complex probabilistic modeling. She actively develops methodologies to address critical, large scale data challenges in genetics, public health, biology, and environmetrics. Supported by major funding from NSERC, CANSSI, and Genome BC, Dr. Wang has published about 60 peer reviewed papers in top-tier statistical journals and machine learning conferences. Website:
https://www.sfu.ca/~lwa68/
