Reproduce “Spatio-Temporal Graph Neural Networks: A Survey” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Zahraa Al Sahili, Mariette Awad
A comprehensive survey categorizes Spatio-Temporal Graph Neural Networks (STGNNs) into hybrid and solo-GNN architectures, detailing their various algorithmic designs and applications across domains like traffic forecasting and human-object interaction. The work also identifies key challenges in automation, benchmarking, and data augmentation for these dynamic models.
Key insights
- Spatio-Temporal GNNs are essential for effectively modeling dynamic, interconnected real-world data where both spatial correlations and temporal dynamics are present.
- STGNN architectures primarily fall into two categories: modular 'Hybrid' designs with separate spatial and temporal components, and 'Solo-GNN' designs that intrinsically integrate temporal information.
- These models demonstrate wide applicability and superior performance in diverse domains such as multivariate time series forecasting and human-object interaction.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.