This repository contains code which accompanies the paper Gromov-Wasserstein meets Weisfeiler-Lehman.
To run the code, you will need the following packages: numpy, POT, networkx, torch-geometric, grakel
Additionally, we use the KSVM implemented in the WTK library and the Wasserstein Weisfeiler-Lehman (WWL) kernel implemented here.
Computations for the Weisfeiler-Lehman distance are included in the utils/distances.py file. Currently, node labels based on degree and size of the graph are supported. The classification experiments with both nearest neighbor and SVMs are included in the experiments folder.