SeeExplainer Captures Synergistic Edge Effects in Graph Neural Networks

SeeExplainer is a parameter-free GNN explainer that uses a granular-ball graph refinement mechanism to decompose graphs into disjoint nodes. This approach captures synergistic effects between edges, addressing the limitations of traditional perturbation-based edge importance methods.