Run the automatic env setup file source setup.sh
or
-
create virtual env (restrict to python=3.7, which is compatible with CPLEX 12.10)
conda create -n ogw python=3.7
-
install pytorch / pytorch-geometric
conda install pytorch=1.8.0 torchvision torchaudio cudatoolkit=11.1 -c pytorch -c nvidia
pip install torch-scatter torch-sparse torch-cluster torch-spline-conv -f https://pytorch-geometric.com/whl/torch-1.8.0+cu111.html
pip install torch-geometric pot
-
install supplimentary libraries
conda install matplotlib joblib networkx numba
-
install the package as develop mode
python setup.py develop
- download MNIST 2D from Kaggle.
- TUDataset benchmark will be loaded from
pyg
.
We provide a set of demonstrations of OGW
:
tightness_syn.ipynb
: A demo of tightness on synthetic data.tightness_mutag.ipynb
: A demo of tightness on MUTAG dataset.barycenter_syn.ipynb
: A demo of barycenter on synthetic data.barycenter_mnist_2d.ipynb
: A demo of barycenter on point cloud MNIST-2D data.
@inproceedings{jin2022orthogonal,
title={Orthogonal Gromov-Wasserstein Discrepancy with Efficient Lower Bound},
author={Jin, Hongwei and Yu, Zishun and Zhang, Xinhua},
booktitle={The 38th Conference on Uncertainty in Artificial Intelligence},
year={2022}
}
The project is under MIT license.