Research map / Time series / Long-term forecasting
RI-Loss: A Learnable Residual-Informed Loss for Time Series Forecasting
Jieting Wang, Xiaolei Shang, Feijiang Li, Furong Peng
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 |
|---|---|---|---|
| train with RI-Loss instead of MSE | Informer · ETTh2 (H=96) | MSE: 3.340 -> 2.299 | Improved |
| train with RI-Loss instead of MSE | DLinear · ETTh2 (H=720) | MSE: 0.696 -> 0.582 | Improved |
| train with RI-Loss instead of MSE | DLinear · Exchange (H=336) | MSE: 0.345 -> 0.265 | Improved |
| train with RI-Loss instead of MSE | DLinear · ILI (H=60) | MSE: 2.410 -> 1.970 | Improved |
| train with RI-Loss instead of MSE | Informer · Exchange (H=720) | MSE: 2.206 -> 2.574 | Hurt |
| train with RI-Loss instead of MSE | Informer · Electricity (H=720) | MSE: 0.341 -> 0.367 | Hurt |
| use HSIC dependence term instead of Pearson correlation (ablation) | DLinear · ETTh2 (H=96) | MSE: 0.305 -> 0.279 | Improved |
| use HSIC dependence term instead of Pearson correlation (ablation) | Informer · ETTh2 (H=96) | MSE: 2.569 -> 2.299 | Improved |
| use HSIC dependence term instead of Pearson correlation (ablation) | DLinear · ETTh2 (H=720) | MSE: 0.714 -> 0.582 | Improved |
| use RI-Loss instead of MAE loss (ablation) | DLinear · ETTh2 (H=720) | MSE: 0.590 -> 0.582 | Improved |
| use RI-Loss instead of MAE loss (ablation) | Informer · ETTh2 (H=720) | MSE: 3.083 -> 3.094 | Hurt |
| train with RI-Loss under input noise (SNR -3dB) | DLinear · ETTh2 H=720 noise -3dB | MSE: 0.693 -> 0.579 | Improved |
Closest papers by meaning
- Patch-wise Structural Loss for Time Series Forecasting (ICML 2025 · similarity 0.90)
- Loss Shaping Constraints for Long-Term Time Series Forecasting (ICML 2024 · similarity 0.89)
- Towards a General Time Series Forecasting Model with Unified Representation and Adaptive Transfer (ICML 2025 · similarity 0.87)
- TimePerceiver: An Encoder-Decoder Framework for Generalized Time-Series Forecasting (NeurIPS 2025 · similarity 0.87)
- OLinear: A Linear Model for Time Series Forecasting in Orthogonally Transformed Domain (NeurIPS 2025 · similarity 0.87)
- Multivariate Time Series Forecasting needs Cross Variable Loss (NeurIPS 2026 · 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.
