Research map / Graph learning / Knowledge and temporal graphs
Easy Begun Is Half Done: Spatial-Temporal Graph Modeling with ST-Curriculum Dropout
Hongjun Wang, Jiyuan Chen, Tong Pan, Zipei Fan, Xuan Song, Renhe Jiang, Lingyu Zhang, Yi Xie, Zhongyi Wang, Boyuan Zhang
This paper tried a common change: Add dropout
Add ST-Curriculum Dropout to Z-GCNETs · METR-LA (15/30/60 min) · MAE: 2.62/2.99/3.32 -> 2.50/2.78/3.05 Improved
What 795 accepted papers report for this change
- 549 improved
- 87 no clear effect
- 12 mixed
- 147 hurt
- Don't Drop Dropout: Optimizing Layer Sparsity for Efficient LLM Training and Inference ICML 2026 · apply layer dropout to whole transformer blocks instead of attention and FFN sub-blocks separately · 20 TPP val loss, 271M · val loss: 2.377 -> 2.367Improved
- Quantifying the Uncertainty of Foundation Models with Singular Value Ensembles ICML 2026 · use MC Dropout · vision and SST-2 · accuracy: 86.3 -> 76.6 on Flowers102Hurt
- Motion Planning in Compressed Representation Spaces ICML 2026 · add nested dropout during training · WOMD guided generation · goal FDE: 4.145 -> 1.104Improved
See this change in detailMatched by the wording of each paper's own result, read by AI from the full text. Papers report gains more often than failures, so this shows what gets published, not the chance a change works.
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 |
|---|---|---|---|
| Add ST-Curriculum Dropout to Z-GCNETs | Z-GCNETs · METR-LA (15/30/60 min) | MAE: 2.62/2.99/3.32 -> 2.50/2.78/3.05 | Improved |
| Add ST-Curriculum Dropout to Z-GCNETs | Z-GCNETs · PeMSD7M (15/30/60 min) | MAE: 1.91/2.35/2.98 -> 1.85/2.21/2.84 | Improved |
| Add ST-Curriculum Dropout to Z-GCNETs | Z-GCNETs · Covid-19 (3/5/7 days) | MAE: 3.27/3.42/3.59 -> 2.82/3.07/3.12 | Improved |
| Add ST-Curriculum Dropout to Z-GCNETs | Z-GCNETs · Crime (3/5/7 days) | MAE: 2.10/2.68/3.19 -> 2.06/2.61/3.12 | Improved |
| Replace ordinary dropout with curriculum-based node selection (ablation) | ST-GCN · METR-LA | MAE: 2.87 -> 2.67 | Improved |
| Add ST-Curriculum Dropout instead of ordinary dropout (ablation) | ST-GCN · METR-LA | MAE: 2.89 -> 2.67 | Improved |
| Replace zeroed dropped-node features with neighbor means (ablation) | ST-GCN with STC-Dropout · METR-LA | MAE: 2.67 -> 2.79 | Hurt |
| Add temporal difficulty scoring alongside spatial scoring (ablation) | SC-Dropout · METR-LA | MAE: 2.76 -> 2.67 | Improved |
| Add spatial difficulty scoring alongside temporal scoring (ablation) | TC-Dropout · METR-LA | MAE: 2.73 -> 2.67 | Improved |
| Add the dynamic weighting policy (ablation) | STC-Dropout without weighting · METR-LA | MAE: 2.70 -> 2.67 | Improved |
| Select difficult nodes first rather than easy nodes first | ST-GCN · METR-LA | MAE: 2.89 -> 3.04 | Hurt |
| Use model predicting loss as the difficulty criterion (ablation) | ST-GCN · METR-LA | MAE: 2.89 -> 2.88 | Improved |
Closest papers by meaning
- Towards Dynamic Spatial-Temporal Graph Learning: A Decoupled Perspective (AAAI 2024 · similarity 0.89)
- Efficient Unlearning for Spatio-temporal Graph (Student Abstract) (AAAI 2025 · similarity 0.89)
- ST-FiT: Inductive Spatial-Temporal Forecasting with Limited Training Data (AAAI 2025 · similarity 0.89)
- Spatio-Temporal Pivotal Graph Neural Networks for Traffic Flow Forecasting (AAAI 2024 · similarity 0.87)
- Efficient Traffic Prediction Through Spatio-Temporal Distillation (AAAI 2025 · similarity 0.87)
- Designing Specialized Two-Dimensional Graph Spectral Filters for Spatial-Temporal Graph Modeling (AAAI 2025 · similarity 0.87)
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.
