Towards explaining graph neural networks via preserving prediction ranking and structural dependency

Youmin Zhang, William K. Cheung, Qun Liu, Guoyin Wang, Lili Yang, Li Liu

Abstract: Highlights•We explain GNNs via preserving prediction ranking, enhancing fidelity of explanation.•We introduce a designed differential ranking loss that guides the optimizing process.•We propose a graph transformation schema to explicitly model edge dependencies.•Experiments and ablation studies on four datasets show explainer’s SOTA performance.