Efficient Neural Common Neighbor for Temporal Graph Link Prediction

Xiaohui Zhang, Yanbo Wang, Xiyuan Wang, Muhan Zhang

Keywords: Temporal graph, Neural common neighbor, Efficient

Abstract:

Temporal graphs are ubiquitous in real-world scenarios, such as social network, trade and transportation. Predicting dynamic links between nodes in a temporal graph is of vital importance. \textcolor{blue}{
Traditional memory-based methods typically leverage the temporal neighborhood of interaction histories to generate node embeddings, which are then aggregated to predict links between source and target nodes. However, these methods primarily focus on learning individual node representations and often neglect the nature of pairwise representation learning aspect. While some recent methods attempt to capture pairwise features, they are less emphasized in large-scale datasets like TGB. Meanwhile, most of these
existing methods tend to suffer from high computational complexity due to the repeated calculation of node embeddings.
}
Motivated by the success of Neural Common Neighbor (NCN) for static graph link prediction, we propose \textbf{TNCN}, a temporal version of NCN for link prediction in temporal graphs. Based on a memory-based backbone instead of traditional static graph neural network, TNCN dynamically updates a temporal neighbor dictionary for each node, and utilizes multi-hop common neighbors between the source and target node to learn a more effective pairwise representation. We validate our model on five large-scale real-world datasets from the Temporal Graph Benchmark (TGB), and find that it achieves new state-of-the-art performance on three of them. Additionally, TNCN demonstrates excellent scalability on large datasets, outperforming popular GNN baselines by up to 6.4 times in speed.

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: 7018