Communication Dans Un Congrès Année : 2020

When Spectral Domain Meets Spatial Domain in Graph Neural Networks

Résumé

Convolutional Graph Neural Networks (Con-vGNNs) are designed either in the spectral domain or in the spatial domain. In this paper, we provide a theoretical framework to analyze these neural networks, by deriving some equivalence of the graph convolution processes, regardless if they are designed in the spatial or the spectral domain. We demonstrate the relevance of the proposed framework by providing a spectral analysis of the most popular ConvGNNs (ChebNet, CayleyNet, GCN and Graph Attention Networks), which allows to explain their performance and shows their limits.

Fichier principal
Vignette du fichier
20.icml.gnn.pdf (933.37 Ko) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-03088374 , version 1 (26-12-2020)

Licence

Identifiants

  • HAL Id : hal-03088374 , version 1

Citer

Muhammet Balcilar, Guillaume Renton, Héroux Pierre, Benoît Gaüzère, Sébastien Adam, et al.. When Spectral Domain Meets Spatial Domain in Graph Neural Networks. Thirty-seventh International Conference on Machine Learning (ICML 2020) - Workshop on Graph Representation Learning and Beyond (GRL+ 2020), Jul 2020, Vienna, Austria. ⟨hal-03088374⟩
513 Consultations
426 Téléchargements

Partager

  • More