Research map / Graph learning / Knowledge and temporal graphs
Spatio-Temporal Graph Scattering Transform
Chao Pan, Siheng Chen, Antonio Ortega
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use separable ST-GST instead of the strongest listed baseline | MSR Action3D · MSR Action3D | accuracy: 87.0 vs 82.2 | Improved |
| use separable rather than joint spatio-temporal scattering (ablation) | ST-GST · MSR Action3D | accuracy: achieving $25\%$ increase in classification accuracy | Improved |
| use a joint strong-product graph design (ablation) | ST-GST · MSR Action3D | accuracy: 87.0 -> 61.7 | Hurt |
| use a joint Kronecker graph design (ablation) | ST-GST · MSR Action3D | accuracy: 87.0 -> 61.0 | Hurt |
| use a joint Cartesian graph design (ablation) | ST-GST · MSR Action3D | accuracy: 87.0 -> 59.1 | Hurt |
| increase temporal scales from 5 to 10 with 5 spatial scales and 3 layers (ablation) | Separable ST-GST · MSR Action3D | accuracy: 73.6 -> 81.4 | Improved |
| increase temporal scales from 10 to 15 with 5 spatial scales and 3 layers (ablation) | Separable ST-GST · MSR Action3D | accuracy: 81.4 -> 85.9 | Improved |
| increase temporal scales from 15 to 20 with 5 spatial scales and 3 layers (ablation) | Separable ST-GST · MSR Action3D | accuracy: 85.9 -> 87.0 | Improved |
| increase layers from 3 to 4 at 5 spatial and 5 temporal scales (ablation) | Separable ST-GST · MSR Action3D | accuracy: 73.6 -> 72.9 | Hurt |
| remove the nonlinear activation (ablation) | ST-GST · skeleton-based human action recognition | accuracy | Hurt |
| use ST-GST rather than ST-GCN when the training ratio is less than 20\% | skeleton-based human action recognition · training ratio below 20\% | accuracy | Improved |
| use ST-GST rather than ST-GCN | skeleton-based human action recognition · classification accuracy and running time | running time | Improved |
Closest papers by meaning
- Designing Specialized Two-Dimensional Graph Spectral Filters for Spatial-Temporal Graph Modeling (AAAI 2025 · similarity 0.89)
- Fully-Connected Spatial-Temporal Graph for Multivariate Time-Series Data (AAAI 2024 · similarity 0.88)
- Spatio-Spectral Graph Neural Networks (NeurIPS 2024 · similarity 0.87)
- Over-squashing in Spatiotemporal Graph Neural Networks (NeurIPS 2025 · similarity 0.86)
- ST-FiT: Inductive Spatial-Temporal Forecasting with Limited Training Data (AAAI 2025 · similarity 0.86)
- Supra-Laplacian Encoding for Transformer on Dynamic Graphs (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
