Implementation of batched Sinkhorn iterations for entropy-regularized Wasserstein loss

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

References in corpus (1)

• Differential Properties of Sinkhorn Approximation for Learning with Wasserstein Distance

Cited by in corpus (1)

• 2D Wasserstein Loss for Robust Facial Landmark Detection