Implementation of batched Sinkhorn iterations for entropy-regularized Wasserstein loss
arXiv:1907.01729v24 citationsHas Code
Originality Synthesis-oriented
AI Analysis
This work offers a practical tool for researchers and practitioners in machine learning, but it is incremental as it focuses on implementation rather than new theoretical or algorithmic contributions.
The authors tackled the implementation of entropy-regularized Wasserstein loss by reviewing Cuturi's method and providing a practical PyTorch implementation, with code made available online.
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