Discussion lead: Brian Ward (CCM)
Title: Efficiently Vectorized MCMC on Modern Accelerators
Introduction: https://arxiv.org/abs/2411.04260, especially section 5.2
Main paper: https://arxiv.org/abs/2503.17405
MCMC has increasingly-well-known difficulties taking advantage of modern compute hardware. These two papers will be used to discuss what those are, how they are frequently avoided in other domains, and a novel way of doing so for MCMC algorithms in practice.