Out-of-core graph training systems for training massive graph neural networks on a single machine.
Capsule: An Out-of-Core Training Mechanism for Colossal GNNs (SIGMOD 2025) Paper PDFCode
SWIFT: Enabling Large-Scale Temporal Graph Learning on a Single Machine (SIGMOD 2026) Paper PDFCode
About This Work
This research line focuses on building efficient graph machine learning systems,
especially out-of-core training mechanisms that enable colossal GNNs and large-scale
temporal graph learning on commodity single-machine settings.