The field of graph neural networks (GNNs) is rapidly evolving, with a focus on improving interpretability and robustness. Recent developments have seen the introduction of novel architectures and techniques, such as the use of entropy-driven approaches and multi-granular attention mechanisms, to better capture complex structural and semantic information in graphs. Additionally, there is a growing emphasis on addressing the challenges of out-of-distribution detection and adversarial attacks in heterogeneous graphs. Noteworthy papers in this area include the proposal of EDEN, a data-centric digraph learning paradigm, and the development of GNN-AID, a framework for graph neural network analysis, interpretation, and defense. These advancements have the potential to significantly improve the performance and reliability of GNNs in a wide range of applications. Notable papers include: EDEN, which proposes a novel data-centric digraph learning paradigm, and GNN-AID, which provides a framework for graph neural network analysis, interpretation, and defense.
Advances in Graph Neural Networks and Interpretability
Sources
Addressing Noise and Stochasticity in Fraud Detection for Service Networks
Toward Data-centric Directed Graph Learning: An Entropy-driven Approach
Wide & Deep Learning for Node Classification
RISE: Radius of Influence based Subgraph Extraction for 3D Molecular Graph Explanation
Robustness questions the interpretability of graph neural networks: what to do?
Framework GNN-AID: Graph Neural Network Analysis Interpretation and Defense
Out-of-Distribution Detection in Heterogeneous Graphs via Energy Propagation
Multi-Granular Attention based Heterogeneous Hypergraph Neural Network
Piecewise Constant Spectral Graph Neural Network