Online Program Home
My Program

Abstract Details

Activity Number: 159 - New Statistical Methods in Phylogenetics
Type: Topic Contributed
Date/Time: Monday, July 30, 2018 : 10:30 AM to 12:20 PM
Sponsor: Biometrics Section
Abstract #328423
Title: Beyond Random-Walk MCMC for Bayesian Phylogenetics
Author(s): Frederick Matsen*
Companies: Fred Hutchinson Cancer Research Center
Keywords: phylogenetics; Bayesian inference; Markov Chain Monte Carlo; Sequential Monte Carlo; Hamiltonian Monte Carlo

Bayesian posterior distributions on phylogenetic trees remain difficult to sample despite decades of effort. The complex discrete and continuous model structure of trees means that recent statistical samplers developed for Euclidean space are not applicable to the phylogenetic case. Thus, we are left with random-walk Markov Chain Monte Carlo with uninformed tree modification proposals. These methods traverse tree space slowly because phylogenetic posteriors are concentrated on a small fraction of the very many possible trees.

In this talk, I will describe our recent work moving beyond random walk MCMC sampling of posterior distributions on phylogenetic trees. This includes (1) an online sequential Monte Carlo algorithm that can update existing posterior inferences when additional sequences arrive, (2) an extension of Hamiltonian Monte Carlo to simultaneously explore discrete and continuous aspects of tree space, and (3) "phylogenetic topographer," a strategy to systematically map out the posterior distribution and provide an approximate posterior weight.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program