Analysis of Dirichlet Energies as Over-smoothing Measures
By: Anna Bison, Alessandro Sperduti
We analyze the distinctions between two functionals often used as over-smoothing measures: the Dirichlet energies induced by the unnormalized graph Laplacian and the normalized graph Laplacian. We demonstrate that the latter fails to satisfy the axiomatic definition of a node-similarity measure proposed by Rusch \textit{et al.} By formalizing fundamental spectral properties of these two definitions, we highlight critical distinctions necessary to select the metric that is spectrally compatible with the GNN architecture, thereby resolving ambiguities in monitoring the dynamics.
Similar Papers
Measuring Over-smoothing beyond Dirichlet energy
Machine Learning (CS)
Finds when AI models get too confused.
Comment on "A Note on Over-Smoothing for Graph Neural Networks"
Machine Learning (CS)
Fixes "fuzzy" computer learning for better results.
Metric spaces of walks and Lipschitz duality on graphs
Machine Learning (CS)
Helps computers learn by watching paths.