Research map / Time series / Long-term forecasting
Correcting Forecasts with Multifactor Neural Attention
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| remove attention sparsity (L1) (ablation) | Our Model · All 20 commodities | MAPE: 20.40 -> 23.69 | Hurt |
| remove soft attention (ablation) | Our Model · All 20 commodities | MAPE: 20.40 -> 33.49 | Hurt |
| remove soft attention (ablation) | Our Model · All 20 commodities | Anomaly %: 5.11 -> 11.05 | Hurt |
| remove attention sparsity (ablation) | Our Model · 5 hardest commodities | MAPE: 34.87 -> 38.66 | Hurt |
| remove soft attention (ablation) | Our Model · 5 hardest commodities | MAPE: 34.87 -> 55.72 | Hurt |
| remove soft attention (ablation) | Our Model · 5 hardest commodities | Anomaly %: 11.56 -> 25.98 | Hurt |
| add attention sparsity to HME gating network | Our Gating Network · All 20 commodities | MAPE: 24.98 -> 24.01 | Helped |
| add attention sparsity to HME gating network | Our Gating Network · 5 hardest commodities | MAPE: 38.29 -> 35.89 | Helped |
| replace gating network with attention on hidden representation | Our Model vs Gating Network · All 20 commodities | MAPE: 24.98 -> 20.40 | Helped |
| use Independent Observation structure with neural network but no attention | Neural Network / - Soft Attention · All 20 commodities | MAPE: 28.27 -> 33.49 | Hurt |
| PCA compression of features for SVR | SVR (RBF Kernel) · All 20 commodities | MAPE: 31.53 -> 31.46 | No clear effect |
| end-to-end GRU time series modeler with sparse attention instead of separate baseline forecast | Our Model · All 20 commodities | MAPE: 20.40 -> 20.28 | Helped |
Closest papers by meaning
- CATS: Enhancing Multivariate Time Series Forecasting by Constructing Auxiliary Time Series as Exogenous Variables (ICML 2024 · similarity 0.85)
- Are Self-Attentions Effective for Time Series Forecasting? (NeurIPS 2024 · similarity 0.84)
- Revisiting Attention for Multivariate Time Series Forecasting (AAAI 2025 · similarity 0.84)
- TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables (NeurIPS 2024 · similarity 0.84)
- WAVE: Weighted Autoregressive Varying Gate for Time Series Forecasting (ICML 2025 · similarity 0.83)
- Revitalizing Multivariate Time Series Forecasting: Learnable Decomposition with Inter-Series Dependencies and Intra-Series Variations Modeling (ICML 2024 · similarity 0.83)
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.