mfouda / IPOT

Implementation of Inexact Proximal point method for Optimal Transport

IPOT

Implementation of Inexact Proximal point method for Optimal Transport from paper "A Fast Proximal Point Method for Computing Exact Wasserstein Distance" (https://arxiv.org/abs/1802.04307).

---

Package dependencies

It requires the following Python packages:

Python 3.5 numpy matplotlib pot 0.4.0 tensorflow 1.1.0

---

Included modules

ipot.py: include function ipot_WD that computes the Wasserstein distance, a function ipot_barycenter that computes the Wasserstein barycenter.

ipot_demo.py: Demo for computing Wasserstein distance using ipot.ipot_WD.

learning_demo.py: 1D demo for learning generative model using ipot.ipot_WD. This file is the only one that needs Tensorflow.

barycenter_demo.py: Demo for computing Wasserstein barycenter.

---

How to run the code

1. To compute Wasserstein distance:

python ipot_demo.py

2. To learn 1D generative model:

python learning_demo.py

3. To compute Wasserstein barycenter:

python barycenter_demo.py

---

License

If you use my code, please cite our paper

@article{xie2018fast,
  title={A Fast Proximal Point Method for Computing Exact Wasserstein Distance},
  author={Xie, Yujia and Wang, Xiangfeng and Wang, Ruijia and Zha, Hongyuan},
  journal={arXiv preprint arXiv:1802.04307},
  year={2018}
}