Bermuda Triangles: GNNs Fail to Detect Simple Topological Structures
Machine Learning
2021-05-04 v1 Artificial Intelligence
Abstract
Most graph neural network architectures work by message-passing node vector embeddings over the adjacency matrix, and it is assumed that they capture graph topology by doing that. We design two synthetic tasks, focusing purely on topological problems -- triangle detection and clique distance -- on which graph neural networks perform surprisingly badly, failing to detect those "bermuda" triangles. Datasets and their generation scripts are publicly available on github.com/FujitsuLaboratories/bermudatriangles and dataset.labs.fujitsu.com.
Cite
@article{arxiv.2105.00134,
title = {Bermuda Triangles: GNNs Fail to Detect Simple Topological Structures},
author = {Arseny Tolmachev and Akira Sakai and Masaru Todoriki and Koji Maruhashi},
journal= {arXiv preprint arXiv:2105.00134},
year = {2021}
}
Comments
ICLR 2021 GTRL Poster Presentation: https://openreview.net/forum?id=Vz_Nl9MSQnu