Research map / Time series / Long-term forecasting
StretchTime: Adaptive Time Series Forecasting via Symplectic Attention
Yubin Kim, Viresh Pati, Jevon Twitty, Vinh Pham, Shihao Yang, Jiecheng Lu
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| replace RoPE with SyPE and adaptive warp | StretchTime Transformer · Synthetic warped AR(1), horizon 720 | MSE: 0.411 -> 0.331 | Improved |
| add channel-mixing MLPs (ablation) | StretchTime on synthetic warped data · Synthetic warped AR(1), avg | MSE: 0.176 -> 0.144 | Improved |
| replace RoPE with SyPE plus warp in the Transformer | 3-layer Transformer · Solar | MSE: 0.214 -> 0.195 | Improved |
| replace RoPE with SyPE plus warp in the Transformer | 3-layer Transformer · PEMS03 | MSE: 0.107 -> 0.094 | Improved |
| add learnable symplectic geometry to fixed RoPE with warp (V4 to V5) (ablation) | StretchTime · PEMS08 avg | MSE: 0.125 -> 0.118 | Improved |
| add learnable symplectic geometry to fixed RoPE with warp (V4 to V5) (ablation) | StretchTime · PEMS04 avg | MSE: 0.090 -> 0.089 | Improved |
| replace fixed RoPE with static-time SyPE (V2 to V3) (ablation) | Transformer without warp · PEMS04 avg | MSE: 0.091 -> 0.091 | No clear effect |
| replace fixed RoPE with static-time SyPE (V2 to V3) (ablation) | Transformer without warp · PEMS08 avg | MSE: 0.129 -> 0.125 | Improved |
| add adaptive warp module without positional geometry (V1) vs fixed RoPE (V2) (ablation) | Transformer · PEMS04 avg | MSE: 0.091 -> 0.093 | Hurt |
| add warp to fixed RoPE (V2 to V4) (ablation) | Transformer with RoPE · PEMS04 avg | MSE: 0.091 -> 0.090 | Improved |
| use StretchTime in place of 4 other deep baselines on large/low-noise sets | StretchTime · ETTh2 avg | MSE: 0.384 vs TimeMixer++ 0.339 | Hurt |
| StretchTime vs OLinear on ECL | StretchTime · ECL avg | MSE: 0.174 vs OLinear 0.159 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- TimeMixer++ — TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting (ICLR 2024) · better on 3 of 6 shared results
- OLinear — OLinear: A Linear Model for Time Series Forecasting in Orthogonally Transformed Domain (NeurIPS 2025) · better on 3 of 5 shared results
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
- PAMod: Modeling Cyclical Shifts via Phase-Amplitude Modulation for Non-stationary Time Series Forecasting (NeurIPS 2026 · similarity 0.87)
- Timer-XL: Long-Context Transformers for Unified Time Series Forecasting (ICLR 2025 · similarity 0.87)
- TimePerceiver: An Encoder-Decoder Framework for Generalized Time-Series Forecasting (NeurIPS 2025 · similarity 0.86)
- TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting (NeurIPS 2025 · similarity 0.86)
- DeformableTST: Transformer for Time Series Forecasting without Over-reliance on Patching (NeurIPS 2024 · similarity 0.86)
- Sequence Complementor: Complementing Transformers for Time Series Forecasting with Learnable Sequences (AAAI 2025 · 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.
