arXiv:2505.12526v2 Announce Type: replace Abstract: Temporal graph networks suffer from irregular supervision in realworld dynamic graphs, as most minibatches contain few labeled events. The lack of labels leads to high-variance gradient updates and, consequently, slow wall-clock convergence. To constructively reduce sparsity, our Moving-Averaged Labels (MAL) assigns soft pseudo-targets based on past supervised signals using a running label distribution while leaving the loss and the model architecture unchanged. Thus, supervision gaps are replaced with informative signals independent of a tem
This paper addresses a known limitation in training temporal graph networks, aligning with the current push for more efficient and robust AI models, especially those dealing with dynamic data.
Improving the efficiency of temporal graph networks has implications for various real-world AI applications, from fraud detection to recommendation systems, where data evolves quickly and supervision can be sparse.
The proposed 'Moving-Averaged Labels' (MAL) method provides a way to reduce data sparsity and stabilize training in temporal GNNs without altering existing model architectures or loss functions.
- · AI researchers and developers
- · GNN-based application designers
- · Companies using real-time dynamic data
More robust and faster-converging temporal graph neural networks (GNNs) become available for deployment.
Improved performance in applications requiring analysis of dynamic relationships over time, such as social network analysis, fraud detection, and drug discovery.
Enhanced adoption of GNNs in industrial settings due to reduced training complexity and improved reliability, potentially accelerating advances in fields like supply chain optimization and personalized medicine.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.LG