Tree Wasserstein distance with weight training

This is the demo code for the paper entitled Approximating 1-Wasserstein Distance with Trees (TMLR 2022)

Note that we used the QuadTree and clustertree implementations of Fixed Support Tree-Sliced Wasserstein Barycenter.

Requirements

Install requirements.

sudo pip install -r requirements.txt

Run

Run example.py

python example.py

Citation

@article{
yamada2022approximating,
title={Approximating 1-Wasserstein Distance with Trees},
author={Makoto Yamada and Yuki Takezawa and Ryoma Sato and Han Bao and Zornitsa Kozareva and Sujith Ravi},
journal={Transactions on Machine Learning Research},
year={2022},
url={https://openreview.net/forum?id=Ig82l87ZVU},
note={}
}

Related papers

• Sho Otao, & Makoto Yamada. A linear time approximation of Wasserstein distance with word embedding selection In EMNLP, 2023.
• Cléa Laouar, Yuki Takezawa, & Makoto Yamada. Large-scale similarity search with Optimal Transport In EMNLP, 2023.

Related Github projects

• Supervised Tree-Wasserstein Distances (ICML 2021)
• Fixed Support Tree-Sliced Wasserstein Barycenter (AISTATS 2022)

Contributors

Name : Makoto Yamada (Okinawa Institute of Science and Technology / Kyoto University) and Yuki Takezawa (Kyoto University)

E-mail : makoto (dot) yamada (at) oist.jp