Implementation of batched Sinkhorn iterations for entropy-regularized Wasserstein loss
Machine Learning
2019-07-08 v2 Machine Learning
Abstract
In this report, we review the calculation of entropy-regularised Wasserstein loss introduced by Cuturi and document a practical implementation in PyTorch. Code is available at https://github.com/t-vi/pytorch-tvmisc/blob/master/wasserstein-distance/Pytorch_Wasserstein.ipynb
Keywords
Cite
@article{arxiv.1907.01729,
title = {Implementation of batched Sinkhorn iterations for entropy-regularized Wasserstein loss},
author = {Thomas Viehmann},
journal= {arXiv preprint arXiv:1907.01729},
year = {2019}
}