Research map / Time series / Long-term forecasting
M2FMoE: Multi-Resolution Multi-View Frequency Mixture-of-Experts for Extreme-Adaptive Time Series Forecasting
Yaohui Huang, Runmin Zou, Yun Wang, Laeeq Aslam, Ruipeng Dong
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 Wavelet-view expert branch (ablation) | M2FMoE · Coyote, 72h | RMSE: 555.641 -> 449.944 | Improved |
| add Fourier-view expert branch (ablation) | M2FMoE · Lexington, 72h | RMSE: 870.735 -> 772.836 | Improved |
| add cross-view shared band splitter (vs uniform splitter) (ablation) | M2FMoE · Lexington, 72h | RMSE: 916.925 -> 772.836 | Improved |
| add cross-view shared band splitter (vs uniform splitter) (ablation) | M2FMoE · Coyote, 72h | RMSE: 541.594 -> 449.944 | Improved |
| add multi-resolution fusion module (ablation) | M2FMoE · Almaden, 72h | RMSE: 59.479 -> 54.120 | Improved |
| add multi-resolution fusion module (ablation) | M2FMoE · Vasona, 72h | RMSE: 19.508 -> 19.571 | Hurt |
| add expert diversity and consistency losses (ablation) | M2FMoE · Coyote, 72h | RMSE: 448.150 -> 449.944 | No clear effect |
| add expert diversity and consistency losses (ablation) | M2FMoE · Lexington, 72h | RMSE: 826.408 -> 772.836 | Improved |
| use M2FMoE without extreme-event labels vs baselines without labels | time series forecasters · five reservoirs, 8h and 72h | RMSE: average improvement of 22.30% | Improved |
| compare M2FMoE (no labels) with label-using MCANN and DAN | extreme-enhanced forecasters · five reservoirs, 8h and 72h | RMSE: average improvement of 9.19% | Improved |
| reduce recent segment length appropriately, versus removing it entirely or making it too long (ablation) | M2FMoE · reservoir datasets | prediction accuracy | Mixed |
| increase number of experts beyond 3 or 4 (ablation) | M2FMoE · reservoir datasets | prediction accuracy | Hurt |
Closest papers by meaning
- Learning from Polar Representation: An Extreme-Adaptive Model for Long-Term Time Series Forecasting (AAAI 2024 · similarity 0.89)
- Towards a General Time Series Forecasting Model with Unified Representation and Adaptive Transfer (ICML 2025 · similarity 0.86)
- Adaptive Multi-Scale Decomposition Framework for Time Series Forecasting (AAAI 2025 · similarity 0.85)
- TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting (ICLR 2024 · similarity 0.85)
- WaveletMixer: A Multi-Resolution Wavelets Based MLP-Mixer for Multivariate Long-Term Time Series Forecasting (AAAI 2025 · similarity 0.85)
- ARM: Refining Multivariate Forecasting with Adaptive Temporal-Contextual Learning (ICLR 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.
