Implementation of the GP with nonparametric covariances via multidimensional convolutions (NP-CGP), from our paper "Nonparametric Gaussian Process Covariances via Multidimensional Convolutions" in AISTATS 2023.
The code was jointly written by magnusross and tomcdonald.
To install, clone the repo, and then run:
conda create --name npcgp_env python=3.9
conda activate npcgp_env
pip install -e .
This will install the CPU version of PyTorch, for the GPU version, follow the instructions here to install the correct version for your GPU.
You can run the model on the UCI data with python uci.py, and run python uci.py -h for help with possible arguments.
@InProceedings{pmlr-v206-mcdonald23a,
title = {Nonparametric Gaussian Process Covariances via Multidimensional Convolutions},
author = {Mcdonald, Thomas M. and Ross, Magnus and Smith, Michael T. and \'Alvarez, Mauricio A.},
booktitle = {Proceedings of The 26th International Conference on Artificial Intelligence and Statistics},
pages = {8279--8293},
year = {2023},
editor = {Ruiz, Francisco and Dy, Jennifer and van de Meent, Jan-Willem},
volume = {206},
series = {Proceedings of Machine Learning Research},
month = {25--27 Apr},
publisher = {PMLR},
url = {https://proceedings.mlr.press/v206/mcdonald23a.html},
}