神经网络尽快在很多领域上效果很好，但是其是“黑盒模型”，无法对预测结果给出解释。图神经网络也有类似的问题。很多论文尝试找出“最有信息量”的节点，边，子图来对图神经网络的预测进行解释，如下图。

感谢Dongsheng Luo同学分类整理了一系列GNN可解释的论文awesome-graph-explainability-papers，如下所示：

  • 最有影响力的论文
  • 最新SOTA论文
  • 2021年论文
  • 2020年论文

链接：https://github.com/flyingdoog/awesome-graph-explainability-papers/blob/main/README.md

对GNN解释性感兴趣的，也可以看看下面的论文解读。

KDD'21 | 如何评估GNN的解释性模型？

ICML 2021 | 针对图神经网络的通用因果解释方法

ICML'21 | 基于子图结构的GNN解释模型

Most Influential

  1. Explainability in graph neural networks: A taxonomic survey. Yuan Hao, Yu Haiyang, Gui Shurui, Ji Shuiwang. ARXIV 2020.

  2. Gnnexplainer: Generating explanations for graph neural networks. Ying Rex, Bourgeois Dylan, You Jiaxuan, Zitnik Marinka, Leskovec Jure. NeurIPS 2019.

  3. Explainability methods for graph convolutional neural networks. Pope Phillip E, Kolouri Soheil, Rostami Mohammad, Martin Charles E, Hoffmann Heiko. CVPR 2019.

  4. Parameterized Explainer for Graph Neural Network. Luo Dongsheng, Cheng Wei, Xu Dongkuan, Yu Wenchao, Zong Bo, Chen Haifeng, Zhang Xiang. NeurIPS 2020.

  5. Xgnn: Towards model-level explanations of graph neural networks. Yuan Hao, Tang Jiliang, Hu Xia, Ji Shuiwang. KDD 2020.

  6. Evaluating Attribution for Graph Neural Networks. Sanchez-Lengeling Benjamin, Wei Jennifer, Lee Brian, Reif Emily, Wang Peter, Qian Wesley, McCloskey Kevin, Colwell  Lucy, Wiltschko Alexander. NeurIPS  2020.

  7. PGM-Explainer: Probabilistic Graphical Model Explanations for Graph Neural Networks. Vu Minh, Thai My T.. NeurIPS  2020.

  8. On Explainability of Graph Neural Networks via Subgraph Explorations. Yuan Hao, Yu Haiyang, Wang Jie, Li Kang, Ji Shuiwang. ICML 2021.

Recent SOTA

  1. Counterfactual Supporting Facts Extraction for Explainable Medical Record Based Diagnosis with Graph Network. Wu Haoran, Chen Wei, Xu Shuang, Xu Bo. NAACL 2021.
  2. When Comparing to Ground Truth is Wrong: On Evaluating GNN Explanation Methods. Faber Lukas, K. Moghaddam Amin, Wattenhofer Roger. KDD 2021.
  3. Counterfactual Graphs for Explainable Classification of Brain Networks. Abrate Carlo, Bonchi Francesco. Proceedings of the th ACM SIGKDD Conference on Knowledge Discovery  Data Mining KDD 2021.
  4. Explainable Subgraph Reasoning for Forecasting on Temporal Knowledge Graphs. Zhen Han, Peng Chen, Yunpu Ma, Volker Tresp. International Conference on Learning Representations ICLR 2021.
  5. Generative Causal Explanations for Graph Neural Networks. Lin Wanyu, Lan Hao, Li Baochun. Proceedings of the th International Conference on Machine Learning ICML 2021.
  6. Improving Molecular Graph Neural Network Explainability with Orthonormalization and Induced Sparsity. Henderson Ryan, Clevert Djork-Arné, Montanari Floriane. Proceedings of the th International Conference on Machine Learning ICML 2021.
  7. Explainable Automated Graph Representation Learning with Hyperparameter Importance. Wang Xin, Fan Shuyi, Kuang Kun, Zhu Wenwu.  ICML 2021.
  8. Higher-order explanations of graph neural networks via relevant walks. Schnake Thomas, Eberle Oliver, Lederer Jonas, Nakajima Shinichi, Schütt Kristof T, Müller Klaus-Robert, Montavon Grégoire. arXiv preprint arXiv:2006.03589 2020.
  9. HENIN: Learning Heterogeneous Neural Interaction Networks for Explainable Cyberbullying Detection on Social Media. Chen, Hsin-Yu and Li, Cheng-Te. EMNLP 2020.

Year 2021

  1. [TPAMI 21] Higher-Order Explanations of Graph Neural Networks via Relevant Walks [paper]

  2. [OpenReview 21] FlowX: Towards Explainable Graph Neural Networks via Message Flows [paper]

  3. [OpenReview 21] Interpreting Graph Neural Networks via Unrevealed Causal Learning [paper]

  4. [OpenReview 21] Explainable GNN-Based Models over Knowledge Graphs

  5. [CIKM 2021] Towards Self-Explainable Graph Neural Network [paper]

  6. [ECML PKDD 2021] GraphSVX: Shapley Value Explanations for Graph Neural Networks [paper]

  7. [IJCNN 21] MEG: Generating Molecular Counterfactual Explanations for Deep Graph Networks  [paper]

  8. [KDD workshop 21] CF-GNNExplainer: Counterfactual Explanations for Graph Neural Networks  [paper]

更多论文请去Git里查看。

Year 2020

  1. [OpenReview 20] A Framework For Differentiable Discovery Of Graph Algorithms
  2. [OpenReview 20] Causal Screening to Interpret Graph Neural Networks

更多论文请去Git里查看。