Source code for the paper
Kai Xu, Akash Srivastava and Charles Sutton. “Variational Russian Roulette for Deep Bayesian Nonparametrics.” International Conference on Machine Learning, 2019.
notebooks/bnpvae.ipynb for the notebook that contains almost everything you would want to reproduce from our paper.
The soure code (in
src) is written in Julia mostly with the Knet.jl deep learning framework. I hope the code is clean enough so that even you never used Julia before would understand.
Some basic instructions for people haven't used Julia before would be added soon.
You may occur errors on running the code if you are using packages with different versions from those I used to write the code. Luckily Julia has its build-in way to run codes in a "virtual environment". The file
Manifest.toml I provide in this repository is for that purpose. Basically, if you already convert the notebook into a script using the provided
nb2script.sh, you can then run the script by
julia --project=REPO_PATH scripts/bnpvae.jl ... where
REPO_PATH is the path to your clone of my repository.