SIG: Self-Interpretable Graph Neural Network for Continuous-time Dynamic Graphs

Lanting Fang, Yulian Yang, Kai Wang, Shanshan Feng, Kaiyu Feng, Jie Gui, Hanning Yuan, Yew-Soon Ong

Keywords: Self-Interpretable, Graph neural network, Continuous-time Dynamic Graph, Causal inference

TL;DR: Self-interpretable GNN for continuous-time dynamic graph; predict future links and extracts a subgraph for casual explanations

Abstract: While graph neural networks have demonstrated potential across various applications, explaining their predictions on dynamic graphs remains largely under-explored. This paper introduces a new research task: self-interpretable GNNs for continuous-time dynamic graphs (CTDGs). We aim to predict future links within dynamic graphs while simultaneously providing causal explanations for these predictions. There are two key challenges: (1) capturing the underlying structural and temporal information that remains consistent across both independent and identically distributed (IID) and out-of-distribution (OOD) data, and (2) efficiently generating high-quality link prediction results and explanations. To tackle these challenges, we propose a novel causal inference model, namely the Independent and Confounded Causal Model (ICCM). ICCM is then integrated into a deep learning architecture that considers both effectiveness and efficiency. Extensive experiments demonstrate that our proposed model significantly outperforms existing methods across link prediction accuracy, explanation quality, and robustness to OOD data. Our code and datasets are anonymously released at https://github.com/2024SIG/SIG.

Primary Area: learning on graphs and other geometries & topologies

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 13224