Deep Gaussian Markov Random Fields and their extensions.
We use JAX for very fast computations. Check out the various notebooks to see what the library offers.
Non-official reimplementations of the following models (work in progress, not everything is included in the library yet):
@inproceedings{siden2020deep,
title={Deep gaussian markov random fields},
author={Sid{\'e}n, Per and Lindsten, Fredrik},
booktitle={International conference on machine learning},
pages={8916--8926},
year={2020},
organization={PMLR}
}
@inproceedings{graph_dgmrf,
author = {Oskarsson, Joel and Sid{\'e}n, Per and Lindsten, Fredrik},
title = {Scalable Deep {G}aussian {M}arkov Random Fields for General Graphs},
booktitle = {Proceedings of the 39th International Conference on Machine Learning},
year = {2022}
}
@article{lippert2023deep,
title={Deep Gaussian Markov Random Fields for Graph-Structured Dynamical Systems},
author={Lippert, Fiona and Kranstauber, Bart and van Loon, E Emiel and Forr{\'e}, Patrick},
journal={arXiv preprint arXiv:2306.08445},
year={2023}
}
Install the latest version with pip
pip install dgmrf
Note that it may be simpler to install jax GPU version with torch CPU only version
The project's documentation is available at https://hgangloff.gitlab.io/dmgrf/index.html
-
First fork the library on Gitlab.
-
Then clone and install the library in development mode with
pip install -e .
- Install pre-commit and run it.
pip install pre-commit
pre-commit install
- Open a merge request once you are done with your changes.
Active: Hugo Gangloff