Research map / Time series / Long-term forecasting
KUMA: A Novel Framework with Koopman Separation and Efficient Multilevel Extraction in Time Series Forecasting
Sijie Xiong, Cheng Tang, Atsushi Shimada
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 Koopman module (KDM) (ablation) | KUMA · 12 datasets averaged | MSE: +3.4% | Improved |
| use input-dependent Koopman operator instead of fixed learned space (ablation) | KUMA · 12 datasets averaged | MSE: 5% -3.4% | Improved |
| add element-wise filtering (EWF) (ablation) | KUMA · 12 datasets averaged | MSE: +3.4% | Improved |
| use linear attention instead of vanilla multi-head attention (ablation) | KUMA · 12 datasets averaged | MSE: 14.7% | Improved |
| use linear attention instead of Mamba blocks (ablation) | KUMA · 12 datasets averaged | MSE: 4.4% | Improved |
| add rotary positional embedding (ablation) | KUMA · 12 datasets averaged | MSE: 1.1% | Improved |
| use 4 levels instead of 2 for PEMS datasets (ablation) | KUMA · PEMS04 | MSE: 0.085 -> 0.084 | Improved |
| vary layer count from 1 to 4 on low-variate datasets (ablation) | KUMA · Exchange | MSE: 0.358 -> 0.355 | No clear effect |
| add Gaussian noise to 5 of 96 input points | KUMA · 12 datasets | MSE and MAE: below 5% | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Koopa — Koopa: Learning Non-stationary Time Series Dynamics with Koopman Predictors (NeurIPS 2023) · better on 11 of 12 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
- $K^2$VAE: A Koopman-Kalman Enhanced Variational AutoEncoder for Probabilistic Time Series Forecasting (ICML 2025 · similarity 0.87)
- Sequence Complementor: Complementing Transformers for Time Series Forecasting with Learnable Sequences (AAAI 2025 · similarity 0.87)
- TimePerceiver: An Encoder-Decoder Framework for Generalized Time-Series Forecasting (NeurIPS 2025 · similarity 0.87)
- SKOLR: Structured Koopman Operator Linear RNN for Time-Series Forecasting (ICML 2025 · similarity 0.87)
- Introducing Spectral Attention for Long-Range Dependency in Time Series Forecasting (NeurIPS 2024 · similarity 0.86)
- KoNODE: Koopman-Driven Neural Ordinary Differential Equations with Evolving Parameters for Time Series Analysis (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.
