Skip to content

Latest commit

 

History

History
19 lines (10 loc) · 766 Bytes

README.md

File metadata and controls

19 lines (10 loc) · 766 Bytes

AVI-when-and-why

Code to reproduce the experiments in the paper: "Amortized Variational Inference: When and Why?" by Charles Margossian and David Blei.

Preprint: https://arxiv.org/abs/2307.11018

Overview

This code can be used to reproduce the results and figures in Section 4 "Numerical Experiments" of the paper. There are four models:

  • linear probabilistic model
  • Nonlinear probabilistic model
  • Bayesian neural network
  • Sawtime series

For each model, run the notebook with the corresponding name. Experimental results are saved in a deliv directory (the first time you run this code, you will need to create this directory). Each notebook also contains code which reads in the experimental results and generates the figures used in the paper.