/unbalanced-ot-functionals

It is a repo which allows to compute all divergences derived from the theory of entropically regularized, unbalanced optimal transport. It relies on a pytorch backend.

Primary LanguagePythonMIT LicenseMIT

Unbalanced Optimal Transport functionals

This repository contains the implementation of the paper Sinkhorn Divergences for Unbalanced Optimal Transport in pytorch. If you use this work for your research, please cite the paper:

@article{sejourne2019sinkhorn,
  title={Sinkhorn Divergences for Unbalanced Optimal Transport},
  author={S{\'e}journ{\'e}, Thibault and Feydy, Jean and Vialard, Fran{\c{c}}ois-Xavier and Trouv{\'e}, Alain and Peyr{\'e}, Gabriel},
  journal={arXiv preprint arXiv:1910.12958},
  year={2019}
}

The repository

This repository allows to compute the entropically regularized optimal transport in both balanced and unbalanced settings, with divergences such as Kullback-Leibler and total variation.

All functionals such as regularized OT, Sinkhorn divergence and maximum mean discrepancy is available in common/functional.py.

See examples/plot_unb_gradient_flows_2D_frame.py for an example.