动态专家模型应对时间序列分布漂移，自适应调整专家结构与选择。
Dynamic TMoE: A Drift-Aware Dynamic Mixture of Experts Framework for Non-Stationary Time Series Forecasting
- 基于MMD检测分布漂移，动态创建和淘汰专家以优化容量。
- 引入时序记忆路由，提升在突发变化下的稳定预测能力。
- 适合处理具有频繁模式突变的工业级时间序列数据。
非平稳时间序列预测面临分布漂移演化的挑战，静态模型难以捕捉此类变化。尽管混合专家（MoE）架构为解耦复杂漂移模式提供了前景，但现有方法受限于固定的专家池和无记忆的路由机制，难以适应突发的模式转变。为此，我们提出动态TMoE框架，在学习阶段统一架构演化与时序连续性。通过最大均值差异（MMD）检测分布漂移，动态实例化异构专家并剪枝冗余专家以优化容量；同时，时序记忆路由利用循环状态与异常存储库，实现无需测试时更新的稳定、上下文感知的专家选择。在九个基准上的实验表明，该方法显著优于现有技术，平均降低10.4% MSE与7.8% MAE。代码已公开于https://github.com/andone-07/Dynamic-TMoE。
原文摘要 · Abstract (English)
Non-stationary time series forecasting is challenged by evolving distribution shifts that static models struggle to capture. While Mixture-of-Experts (MoE) architectures offer a promising paradigm for decoupling complex drift patterns, existing approaches are limited by fixed expert pools and memoryless routing, hampering their ability to adapt to abrupt regime shifts. To address this, we propose Dynamic TMoE, a framework that unifies architectural evolution with temporal continuity during learning phase. By detecting distribution shifts via Maximum Mean Discrepancy (MMD), we dynamically instantiate heterogeneous experts and prune redundant ones to optimize capacity. Additionally, a temporal memory router leverages recurrent states and an anomaly repository to ensure stable, context-aware expert selection without requiring test-time updates. Experiments on nine benchmarks demonstrate state-of-the-art performance, reducing MSE by 10.4% and MAE by 7.8%. Code is available at https://github.com/andone-07/Dynamic-TMoE.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。