Network Analysis and Graph Neural Networks

The field of network analysis and graph neural networks is moving towards a deeper understanding of complex network structures and the development of more sophisticated models to analyze and learn from these networks. Researchers are exploring new methods to identify critical edges and nodes in networks, as well as developing more accurate models to predict network behavior and evolution. A key area of focus is the study of homophily and heterophily in networks, which has important implications for understanding social networks and designing effective graph neural network architectures. Notable papers in this area include:

  • A Graph-Neural-Network-Entropy model, which proposes a novel method for vital node identification in networks.
  • ViG-LRGC, which introduces a learnable reparameterized graph construction method for vision graph neural networks.
  • Exploring Heterophily in Graph-level Tasks, which presents a theoretical analysis of heterophily in graph-level learning and offers guidance for designing effective GNN architectures.

Sources

Sensitivity of Perron and Fiedler eigenpairs to structural perturbations of a network

Homophily in Complex Networks: Measures, Models, and Applications

A Graph-Neural-Network-Entropy model of vital node identification on network attack and propagation

ViG-LRGC: Vision Graph Neural Networks with Learnable Reparameterized Graph Construction

Exploring Heterophily in Graph-level Tasks

Graph Neural Networks with Similarity-Navigated Probabilistic Feature Copying

Built with on top of