Research map / Time series / Causal and generative time series
Dynamic TMoE: A Drift-Aware Dynamic Mixture of Experts Framework for Non-Stationary Time Series Forecasting
Jiawen Zhu, Shuhan Liu, Di Weng, Yingcai Wu
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 drift-aware adaptation (ablation) | Dynamic TMoE · ETTh1 | MSE: 0.436 -> 0.429 | Improved |
| add drift-aware adaptation (ablation) | Dynamic TMoE · Weather | MSE: 0.246 -> 0.240 | Improved |
| use GRU temporal router instead of non-temporal routers (ablation) | Dynamic TMoE · six benchmarks | MSE: MSE +4.2% and MAE +1.8% when replaced | Improved |
| add anomaly memory repository (ablation) | Dynamic TMoE · ETTh1 | MSE: 0.438 -> 0.429 | Improved |
| use heterogeneous experts instead of homogeneous (ablation) | Dynamic TMoE · six benchmarks | MSE: MSE +4.5% and MAE +2.1% when homogeneous | Improved |
| add cyclic relation layer (ablation) | Dynamic TMoE · six benchmarks | MSE: MSE +4.5% and MAE +1.8% when removed | Improved |
| use MMD drift detection instead of mean-variance (ablation) | Dynamic TMoE · Exchange | MSE: 0.382 -> 0.351 | Improved |
| use post-addition alignment instead of fine-tuning base experts (ablation) | Dynamic TMoE · ILI | MSE: 2.238 -> 1.981 | Improved |
| use Drift Pattern Profiler instead of random expert type (ablation) | Dynamic TMoE · ILI | MSE: 2.260 -> 1.981 | Improved |
| add expert pruning (ablation) | Dynamic TMoE · Exchange | MSE: 0.355 -> 0.351 | Improved |
| increase top-k from 3 to 4 (ablation) | Dynamic TMoE · four datasets | MSE | Hurt |
| use top-k=1 instead of 3 (ablation) | Dynamic TMoE · four datasets | MSE | Hurt |
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.
- TimeMixer — TimeMixer: Decomposable Multiscale Mixing for Time Series Forecasting (ICLR 2024) · better on 5 of 6 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
- Learning Pattern-Specific Experts for Time Series Forecasting Under Patch-level Distribution Shift (NeurIPS 2025 · similarity 0.88)
- TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting (NeurIPS 2025 · similarity 0.87)
- Mixture of Online and Offline Experts for Non-Stationary Time Series (AAAI 2025 · similarity 0.86)
- Slimming the Fat-Tail: Morphing-Flow for Adaptive Time Series Modeling (ICML 2025 · similarity 0.86)
- Kairos: Toward Adaptive and Parameter-Efficient Time Series Foundation Models (NeurIPS 2026 · similarity 0.86)
- SynTSBench: Rethinking Temporal Pattern Learning in Deep Learning Models for Time Series (NeurIPS 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.
