Quantile-Propagation-for-Wasserstein-Approximate-Gaussian-Processes

This repo contains a demo for the NeurIPS-2020 publication "Quantile Propagation for Wasserstein-Approximate Gaussian Processes". To get the results reported in the paper, please run the code in the branch "no-lookup-table", which uses no lookup table but exact computation.

Steps to run the code

1. Install virtual environment: conda create -n QP python=3.6
2. Activate environment: conda activate QP
3. Install requirements: pip install -r requirements.txt
4. Download lookup tables from google drive to [the repo path]/data/
5. Enter the experiment dir: cd [the repo path]/experiments/
6. run experiments: python classification.py

Citation

If you find Quantile Propagation for Wasserstein-Approximate Gaussian Processes useful in your research, please consider citing:

@article{zhang2020wassapproxgp,
	title={Quantile Propagation for Wasserstein-Approximate Gaussian Processes},
	author={Zhang, Rui and Walder, Christian J. and Bonilla, Edwin V. and Rizoiu, Marian-Andrei and Xie, Lexing},
	journal={the 34th Conference on Neural Information Processing Systems (NeurIPS 2020)},
	year={2020}
}

License

MIT License