I am happy to present Alexandre Bouchard in the SPG seminar series.
Place: Floor 2 lunchroom of Science for Life Laboratory
Date: Friday, April 19
Title: Phylogenetic Sequential Monte Carlo
Abstract: Bayesian phylogenetic methods rely on accurate approximations of intractable expectations over the space of phylogenies, a problem that has been traditionally approached almost exclusively using Markov chain Monte Carlo samplers. However, the advent of large scale phylogenetic problems and new high resolution models calls for complementary approaches. In this talk, I will describe one such complementary approach based on Sequential Monte Carlo methods. I will demonstrate the flexibility that can be achieved using extensions of standard particle filters, and show different applications we are currently working on. I will also describe a new method that we have developed to parallelize SMC algorithms, entangled Monte Carlo. This technique, based on the particle genealogy, can be applied to parallelize SMC calculations while keeping communication costs between nodes low.