NODE-SAT: Temporal Graph Learning with Neural ODE-Guided Self-Attention
Keywords: Temporal Graph, Neural ODE, Link prediction
Abstract: We propose NODE-SAT, a novel temporal graph learning model that integrates Neural Ordinary Differential Equations (NODEs) with self-attention mechanisms. NODE-SAT's design requires only historical 1-hop neighbors as input and comprises three key components: a temporal link processing module utilizing NODE-guided self-attention layers to capture temporal link information, a node representation module summarizing neighbor information, and a prediction layer. Extensive experiments across thirteen temporal link prediction datasets demonstrate that NODE-SAT achieves state-of-the-art performance on most datasets with significantly faster convergence. The model demonstrates high accuracy, rapid convergence, robustness across varying dataset complexities, and strong generalization capabilities in both transductive and inductive settings in temporal link prediction. These findings highlight NODE-SAT's effectiveness in capturing node correlations and temporal link dynamics.
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.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
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: 13384
Loading