Higher-order clustering and pooling for GNNs
Paper accepted at the 31st ACM International Conference on Information and Knowledge Management (CIKM, 2022). https://arxiv.org/abs/2209.03473
In this repo, we provide: (1) the code for the pooling operator itself in hoscpool.py. It can be used similarly to MinCutPool or DiffPool and will soon be available in Pytorch Geometric. (2) a short example of how to use hoscpool for an arbitrary graph classification task.
If needed, install the required packages: torch and torch-geometric.
Please cite the original paper if you are using HoscPool in your work.
@inproceedings{duval2022higher,
title={Higher-order clustering and pooling for graph neural networks},
author={Duval, Alexandre and Malliaros, Fragkiskos},
booktitle={Proceedings of the 31st ACM International Conference on Information \& Knowledge Management},
pages={426--435},
year={2022}
}