Research map / Time series / Long-term forecasting
Sonnet: Spectral Operator Neural Network for Multivariable Time Series Forecasting
Yuxuan Shu, Vasileios Lampos
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 naive attention with MVCA (ablation) | PatchTST, Samformer, iTransformer · ILI tasks | MAE: -10.7% MAE on average | Improved |
| replace naive attention with MVCA (ablation) | PatchTST · ILI-US2 H=28 | MAE: 1.1525 -> 0.8765 | Improved |
| remove attention module from base models (ablation) | iTransformer/Samformer/PatchTST · ILI tasks | MAE | No clear effect |
| replace naive attention with FEDformer FED attention (ablation) | base models · ILI tasks | MAE: e.g. PatchTST 1.1525 -> 1.4953 (ILI-US2 H=28) | Hurt |
| replace naive attention with FNet attention (ablation) | base models · ILI tasks | MAE | Hurt |
| use MVCA instead of VDAB attention (ablation) | base models · ILI tasks | MAE: -3.5% MAE | Improved |
| use MVCA instead of VDAB inside Sonnet (ablation) | Sonnet · ILI and WEA (32 tasks) | MAE: -3.6% MAE on average | Improved |
| use MVCA instead of naive attention inside Sonnet (ablation) | Sonnet · ILI and WEA (32 tasks) | MAE: -7% MAE on average | Improved |
| use MVCA instead of FED attention inside Sonnet (ablation) | Sonnet · ILI and WEA (32 tasks) | MAE: -51.5% MAE | Improved |
| use Sonnet instead of DeformTime | Sonnet · 47 tasks overall | MAE: -2.2% MAE on average (best on 34 of 47) | Improved |
| use Sonnet instead of baselines on ETT | Sonnet · ETTh1/ETTh2 | MAE: ETTh2 H=720: 0.3640 -> 0.4335 | Hurt |
| use Sonnet on tasks other than ETT | Sonnet · non-ETT tasks | MAE: -3.3% MAE on average | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- TimeXer — TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables (NeurIPS 2024) · 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
- Introducing Spectral Attention for Long-Range Dependency in Time Series Forecasting (NeurIPS 2024 · similarity 0.88)
- Revisiting Attention for Multivariate Time Series Forecasting (AAAI 2025 · similarity 0.87)
- OLinear: A Linear Model for Time Series Forecasting in Orthogonally Transformed Domain (NeurIPS 2025 · similarity 0.86)
- HDT: Hierarchical Discrete Transformer for Multivariate Time Series Forecasting (AAAI 2025 · similarity 0.86)
- Temporal Query Network for Efficient Multivariate Time Series Forecasting (ICML 2025 · similarity 0.86)
- Shifting Time: Time-series Forecasting with Khatri-Rao Neural Operators (ICML 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.
