Mathieu Fourment

So glad to share this work on Bayesian phylogenetics in hyperbolic space! Hyperbolic embeddings change trees from discrete objects into continuous parameters, giving “nearby” trees. arxiv.org/abs/2206.08044

Great opportunity to do a PhD on a relevant issue. 2 PhD scholarships available to develop realtime outbreak surveillance methods. Join @koadman and myself @ithreeinst. bit.ly/2WM9Skv and bit.ly/2LhnPS6

Tomorrow Cheng Zhang (Peking University) will be describing his work on variational Bayesian phylogenetic inference. Tuesday, November 12, 4:00 PM PST youtube.com/watch?v=8ZnH2d…

@avehtari @koadman @mcmc_stan Good eye. We don't mention it in the preprint but the relative tolerance we used was 0.001. As you said, the default value (0.01) was definitely too high.

Today, most Bayesian phylogenetics uses MCMC. But there are other, potentially much faster ways to do Bayesian inference. In a new manuscript @koadman and I explore variational inference in the Stan language as an alternative biorxiv.org/content/10.110…