Research map / Time series / Long-term forecasting
Gaussian Adaptive Patching Powered Fully-Connected Spatial-Temporal Graph for Multivariate Time-Series Data
Yue Wang, Min Wu, Yuecong Xu, Jianfei Yang, Xiaoli Li, Lihua Xie, Zhenghua Chen
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 FC graph construction to per-patch baseline (ablation) | baseline GNN · C-MAPSS FD001 | RMSE: 12.58±0.22 -> 12.17±0.16 | Improved |
| add moving-window FC graph convolution (no pooling) (ablation) | BL + FC Cons. · C-MAPSS FD001 | RMSE: 12.17±0.16 -> 12.03±0.22 | Improved |
| add temporal pooling to FC graph convolution (ablation) | BL + FC Cons. + Conv. (w/o Pool) · C-MAPSS FD001 | RMSE: 12.03±0.22 -> 11.62±0.19 | Improved |
| add decay matrix to FC graph (ablation) | GAP-STGNN · C-MAPSS FD001 | RMSE: 12.15±0.17 -> 11.44±0.42 | Improved |
| add threshold truncation to Gaussian receptive fields (ablation) | FC-STGNN + Gau. · C-MAPSS FD001 | RMSE: 11.84±0.33 -> 11.49±0.33 | Improved |
| add adaptive patch selection (ablation) | FC-STGNN + Gau. · C-MAPSS FD001 and FD004 | RMSE: 11.49±0.33 -> 11.44±0.42 (FD001); 13.47±0.22 -> 13.51±0.32 (FD004) | Mixed |
| soft attenuation of less relevant patches instead of discarding (ablation) | GAP-STGNN · UCI-HAR | accuracy: 94.01±2.18 -> 96.31±0.18 | Improved |
| add decay matrix to FC graph (ablation) | GAP-STGNN · UCI-HAR | accuracy: 95.20±0.27 -> 96.31±0.18 | Improved |
| add Gaussian adaptive patching to FC-STGNN | FC-STGNN · C-MAPSS FD002 | Score: 738±49 -> 691±56 | Improved |
| add Gaussian adaptive patching to FC-STGNN | FC-STGNN · SRSCP1 | accuracy: 90.93±0.72 -> 93.04±0.51 | Improved |
| increase moving window size from 1 to 2 (ablation) | FC-STGNN · C-MAPSS | RMSE | Improved |
| set decay rate to 0.1 (ablation) | FC-STGNN · C-MAPSS and classification | RMSE/accuracy | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- TSLANet — TSLANet: Rethinking Transformers for Time Series Representation Learning (ICML 2024) · better on 1 of 1 shared result
- TimesNet — TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis (ICLR 2023) · better on 1 of 1 shared result
- TS2Vec — TS2Vec: Towards Universal Representation of Time Series (AAAI 2022) · better on 0 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Fully-Connected Spatial-Temporal Graph for Multivariate Time-Series Data (AAAI 2024 · similarity 0.95)
- ST-FiT: Inductive Spatial-Temporal Forecasting with Limited Training Data (AAAI 2025 · similarity 0.88)
- Irregular Multivariate Time Series Forecasting: A Transformable Patching Graph Neural Networks Approach (ICML 2024 · similarity 0.88)
- Hi-Patch: Hierarchical Patch GNN for Irregular Multivariate Time Series (ICML 2025 · similarity 0.88)
- Towards Dynamic Spatial-Temporal Graph Learning: A Decoupled Perspective (AAAI 2024 · similarity 0.88)
- Designing Specialized Two-Dimensional Graph Spectral Filters for Spatial-Temporal Graph Modeling (AAAI 2025 · similarity 0.88)
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.
