Created: November 06, 2020
Modified: April 26, 2022
Modified: April 26, 2022
variational inference
This page is from my personal notes, and has not been specifically reviewed for public consumption. It might be incomplete, wrong, outdated, or stupid. Caveat lector.How should people do VI?
- One ultimate goal is that you write a Stan model (or better, a model with discrete variables, but one step at a time) and inference 'just works', quickly, reliably, and accurately.
- Another ultimate goal is to use VI to extend the classes of models we can fit. That includes S(N)LDS, phylogenetic trees, maybe other stuff.
The most important question for the Stan stuff is probably automatic surrogate building. If we can build good general-purpose surrogates, we're doing pretty well.