Implement NUTS with multinomial sampling and generalised U-Turn criterion #1648
ben18785
started this conversation in
New methods
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
This is different to #293 and is based on the musings here. This is actually what Stan runs on and should be its own thing (potentially called "XHMC'' as in the Betancourt paper).
The Stan 2.17 manual says, "Rather than using a standard Metropolis step, the final parameter value is selected via multinomial sampling with a bias toward the second half of the steps in the trajectory."
There's a discussion of the advances here. Probably the best description of the algorithmic changes are given on page 52+ in this intro paper.
Beta Was this translation helpful? Give feedback.
All reactions