This repository contains code for the paper:

Probabilistic Circuits for Variational Inference in Discrete Graphical Models

"Probabilistic Circuits for Variational Inference in Discrete Graphical Models"
Andy Shih, Stefano Ermon
In Proceedings of the 34th Conference on Neural Information Processing Systems (NeurIPS), 2020

@inproceedings{ShihEneurips20,
  author    = {Andy Shih and Stefano Ermon},
  title     = {Probabilistic Circuits for Variational Inference in Discrete Graphical Models},
  booktitle = {Advances in Neural Information Processing Systems 33 (NeurIPS)},
  month     = {december},
  year      = {2020},
}

Here are commands for running the experiments:

Ising Models

python runising.py --loadgm=1000 --run=123 --mode=2 --n=4
python runising.py --loadgm=1000 --run=123 --mode=1 --n=8
python runising.py --loadgm=1000 --run=123 --mode=1 --n=16
python runising.py --loadgm=1000 --run=123 --mode=1 --n=32

, and repeat with loadgm=[1001,1002,1003].

UAI Inference Competition

python runuai.py --run=123

Contact

For questions, contact us at:

andyshih at cs dot stanford dot edu