Talk on multi scale delayed rejection and the future of Stan's samplers

I gave a talk recently in the University of Michigan Science and ML seminar series, organized by Keith Phuthi. Sam Livingstone was kind enough to moderate. We’re very excited about the mixture of delayed rejection and partial momentum refresh one-step HMC. Here are the slides:

sciml2023.pdf (5.6 MB)

and here’s the recording of the talk: