On Incorporating Scale into Graph Networks

Christian Koke, Yuesong Shen, Abhishek Saroha, Marvin Eisenberger, Bastian Rieck, Michael M. Bronstein, Daniel Cremers

Track: New scientific result

Keywords: Generalization, (Resolution-)Scale, Graph Neural Networks

TL;DR: Standard Graph Neural Networks turn out to not be continuous, which precludes them from consistently incorporating varying (resolution-)scale information.

Abstract: Standard graph neural networks assign vastly different latent embeddings to graphs describing the same physical system at different resolution scales. This precludes consistency in applications and prevents generalization between scales as would fundamentally be needed in many scientific applications. We uncover the underlying obstruction, investigate its origin and show how to overcome it.

Presenter: ~Christian_Koke1

Submission Number: 12