Research map / Graph learning / Knowledge and temporal graphs
Recurrent Space-time Graph Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add RSTG to C2D | C2D with 2D ResNet-50 · Something-Something-v1 val | Top-1 accuracy: 31.7 -> 42.8 | Helped |
| Add RSTG to C2D | C2D with 2D ResNet-50 · Something-Something-v1 val | Top-5 accuracy: 64.7 -> 73.6 | Helped |
| Add RSTG to I3D | I3D with 3D ResNet-50 · Something-Something-v1 val | Top-1 accuracy: 41.6 -> 49.2 | Helped |
| Add RSTG to I3D | I3D with 3D ResNet-50 · Something-Something-v1 val | Top-5 accuracy: 72.2 -> 78.8 | Helped |
| Use sparse rather than fully connected graph adjacency (ablation) | RSTG-to-vec · Something-Something-v1 | Top-1 accuracy: 46.9 -> 47.7 | Helped |
| Use sparse rather than fully connected graph adjacency (ablation) | RSTG-to-vec · Something-Something-v1 | Top-5 accuracy: 76.8 -> 77.9 | Helped |
| Use graph processing on higher-level res4 features instead of lower-level res2 features (ablation) | RSTG with I3D backbone · Something-Something-v1 | accuracy | Helped |
| Stack two graphs at res3 and res4 (ablation) | RSTG with I3D backbone · Something-Something-v1 | accuracy | Helped |
| Remove temporal recurrence and average-pool each node across time (ablation) | Space-Only RSTG · SyncMNIST | accuracy | Hurt |
| Remove spatial message passing between nodes (ablation) | Time-Only RSTG · SyncMNIST | accuracy | Hurt |
| Share parameters between spatial and temporal MLPs (ablation) | Homogeneous Space-time RSTG · SyncMNIST | accuracy | Hurt |
| Schedule three spatial iterations before one final temporal update (ablation) | 1-temp RSTG · SyncMNIST | accuracy | Hurt |
Closest papers by meaning
- Structural-RNN: Deep Learning on Spatio-Temporal Graphs (CVPR 2016 · similarity 0.89)
- Discovering Dynamic Salient Regions for Spatio-Temporal Graph Neural Networks (NeurIPS 2021 · similarity 0.88)
- Learning to Evolve Scenes: Reasoning about Human Activities with Scene Graphs (NeurIPS 2026 · similarity 0.87)
- State Space Models on Temporal Graphs: A First-Principles Study (NeurIPS 2024 · similarity 0.86)
- Over-squashing in Spatiotemporal Graph Neural Networks (NeurIPS 2025 · similarity 0.86)
- Efficient Unlearning for Spatio-temporal Graph (Student Abstract) (AAAI 2025 · similarity 0.85)
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.