#machine learning
May 2025
Graph Representational Learning: When Does More Expressivity Hurt Generalization?
A family of premetrics that capture different degrees of structural similarity between graphs are introduced and relate these similarities to generalization, and consequently, the performance of expressive GNNs are related.
Sohir Maskey, Raffaele Paolino, Fabian Jogl et al.
· arXiv.org · 4 citations
· ⚡1