Closed FrankD closed 4 years ago
Frank, and ESS implementation would be awesome, thanks! BGR optional, but ESS might be useful in-chain as well as post-hoc.
Disappointingly (for my coding enjoyment), it turns out that this is already implemented in Tensorflow probability.
I added calculation of the ESS to the end of the mcmc.py
script. Pull request #7 (but it's literally one line).
@chrism0dwk, I was thinking of implementing some convergence diagnostics for your MCMC; is this useful? I'm thinking effective sample size and maybe Gelman-Rubin R-hat.
https://arxiv.org/pdf/1909.11827.pdf