On the Reproducibility of “G-Mixup:Graph Data Augmentation for Graph Classification”

Dylan Cordaro, Shelby Cox, Yiman Ren, Teresa Yu

Keywords: Python, graphical neural networks, mixup, graphon

Abstract: We attempt to reproduce results on a novel graph data mixup method called "G-Mixup". We investigate both theoretical and experimental claims on the effectiveness of G-Mixup to produce mixed-up synthetic graphs and to improve the performance of graph neural networks. Although we are able to reproduce some of the original paper's experimental results to within 6% of their reported values, our results do not provide statistically significantly evidence that G-Mixup improves the performance of GNNs, nor does it perform better than other data augmentation methods.

Paper Url: https://proceedings.mlr.press/v162/han22c.html

Paper Venue: ICML 2022

Confirmation: The report pdf is generated from the provided camera ready Google Colab script, The report metadata is verified from the camera ready Google Colab script, The report contains correct author information., The report contains link to code and SWH metadata., The report follows the ReScience latex style guides as in the Reproducibility Report Template (https://paperswithcode.com/rc2022/registration)., The report contains the Reproducibility Summary in the first page., The latex .zip file is verified from the camera ready Google Colab script

Journal: ReScience Volume 9 Issue 2 Article 34

Doi: https://www.doi.org/10.5281/zenodo.8173737

Code: https://archive.softwareheritage.org/swh:1:dir:9c61eef8a6a063061f952ef48145dd9389d45281