Research map / Time series / Long-term forecasting
Topological Attention for Time Series Forecasting
Sebastian Zeng, Florian Graf, Christoph Hofer, Roland Kwitt
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 |
|---|---|---|---|
| add topological attention to N-BEATS | N-BEATS · M4 average | OWA: 0.814 -> 0.811 | Improved |
| add topological attention to N-BEATS on Others subgroup | N-BEATS · M4 Others | sMAPE: 3.789 -> 3.803 | Hurt |
| add topological attention to N-BEATS on Yearly subgroup | N-BEATS · M4 Yearly | sMAPE: 13.149 -> 13.063 | Improved |
| add topological features without attention (+Top) to N-BEATS (ablation) | N-BEATS · M4 ablation (T=2H) | OWA: 0.827 -> 0.920 | Hurt |
| add attention on raw windows (+Attn) to N-BEATS (ablation) | N-BEATS · M4 ablation (T=2H) | sMAPE: 11.488 -> 11.492 | No clear effect |
| add full topological attention to N-BEATS (ablation) | N-BEATS · M4 ablation (T=2H) | sMAPE: 11.488 -> 11.466 | Improved |
| add topological features without attention (+Top) to linear model (ablation) | Linear model · electricity single series | average rank (sMAPE): 2.40 -> 3.40 | Hurt |
| add topological attention to linear model (ablation) | Linear model · electricity single series | average rank (sMAPE): 2.40 -> 1.90 | Improved |
| add topological attention to linear model (ablation) | Linear model · car-parts single series | average rank (sMAPE): 2.75 -> 2.50 | Mixed |
| add attention on raw windows (+Attn) to linear model (ablation) | Linear model · car-parts single series | average rank (sMAPE): 2.75 -> 1.25 | Improved |
Closest papers by meaning
- Time-Aware Knowledge Representations of Dynamic Objects with Multidimensional Persistence (AAAI 2024 · similarity 0.87)
- Introducing Spectral Attention for Long-Range Dependency in Time Series Forecasting (NeurIPS 2024 · similarity 0.86)
- TimeBase: The Power of Minimalism in Efficient Long-term Time Series Forecasting (ICML 2025 · similarity 0.86)
- Revisiting Attention for Multivariate Time Series Forecasting (AAAI 2025 · similarity 0.85)
- Enhancing Foundation Models for Time Series Forecasting via Wavelet-based Tokenization (ICML 2025 · similarity 0.85)
- CATS: Enhancing Multivariate Time Series Forecasting by Constructing Auxiliary Time Series as Exogenous Variables (ICML 2024 · similarity 0.85)
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.
