On reproduction of On the regularization of Wasserstein GANs

TL;DR

This paper investigates the reproducibility of the 2018 Wasserstein GAN regularization paper, focusing on key experimental aspects like learning speed, stability, and sampling methods, providing open source code for validation.

Contribution

It systematically reproduces and evaluates the original paper's experiments, clarifying which results are reproducible and at what resource cost.

Findings

Reproduced learning speed and stability results

Assessed robustness against hyperparameters

Evaluated Wasserstein distance estimation methods

Abstract

This report has several purposes. First, our report is written to investigate the reproducibility of the submitted paper On the regularization of Wasserstein GANs (2018). Second, among the experiments performed in the submitted paper, five aspects were emphasized and reproduced: learning speed, stability, robustness against hyperparameter, estimating the Wasserstein distance, and various sampling method. Finally, we identify which parts of the contribution can be reproduced, and at what cost in terms of resources. All source code for reproduction is open to the public.