Two papers describing Bayes4Health funded research have been accepted for AISTATS 2022.

The first “A predictive approach to Bayesian nonparametric survival analysis” by Edwin Fong and Brieuc Lehmann develops a new approach to Bayesian survival analysis based on a recasting of Bayesian inference in terms of assigning a predictive distribution on the unseen values of a population conditional on the observed data. The advantage of such an approach is that it avoids the need to specify a prior — which can be challenging to elicit — and can lead to simpler and more scalable fitting of the model.

The second “Efficient computation of the volume of a polytope in high-dimensions using piecewise deterministic Markov processes” by Augustin Chevallier, Frederic Cazals and Paul Fearnhead shows how recent advances in MCMC can be used to efficiently calculate volumes of polytopes. This is an important application, across many disciplines, and is related to the problem of calculating marginal likelihoods in Bayesian statistics. The new method uses MCMC based on simulating piecewise deterministic Markov processes. These methods share the excellent mixing of Hamiltonian Monte Carlo methods for this application, but with the important advantage of much more efficient implementation: which can lead to orders of magnitude improvements in terms of overall efficiency.