Boosting Temporal Graph Learning From Global and Local Perspectives

Fengyi Wang, Guanghui Zhu, Chunfeng Yuan, Yihua Huang

Primary Area: general machine learning (i.e., none of the above)

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.

Keywords: temporal graphs, graph neural networks, attention mechanism

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

TL;DR: We present that modeling from global and local viewpoints is indispensable for temporal graph representation learning, and propose the Global and Local Embedding Network(GLEN) to effectively generate node embeddings by considering both perspectives.

Abstract: Extensive research has been dedicated to learning on temporal graphs due to its wide range of applications. Some works intuitively merge GNNs and RNNs to capture structural and temporal information, while recent works propose to aggregate information from neighbor nodes in local subgraphs based on message passing or random walk. These methods produce node embeddings from a global or local perspective and ignore the complementarity between them, thus facing limitations in capturing complex and entangled dynamic patterns when applied to diverse datasets or evaluated by more challenging evaluation protocols. To address the challenges, we propose the Global and Local Embedding Network (GLEN) for effective and efficient temporal graph representation learning. Specifically, GLEN dynamically generates embeddings for graph nodes by considering both global and local perspectives. Then, global and local embeddings are elegantly combined by a cross-perspective fusion module to extract high-order semantic relations in graphs. We evaluate GLEN on multiple real-world datasets and apply several negative sampling strategies. Sufficient experimental results demonstrate that GLEN outperforms other baselines in both link prediction and dynamic node classification tasks.

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: 1741