提出动态学习新视角，提升时间序列预测模型性能
Time Series Forecasting Through the Lens of Dynamics
- 从动态演化角度分析模型，识别出关键的动态学习能力
- 发现模型末端设置动态模块能显著提升预测效果
- 提供可即插即用的方法，适合改进各类时序模型
尽管深度学习在多模态中趋于同质化，但其在时间序列预测任务中仍逊于浅层线性模型。我们假设模型应直接学习从过去到未来的数据演化规律，这种能力称为动态学习能力。为此，我们提出原创的 $ exttt{PRO-DYN}$ 框架，用于分析现有模型的动态特性。研究发现：1）表现不佳的架构仅部分学习动态；2）动态模块置于模型末尾至关重要。通过系统性和实证研究，我们在多种骨干网络、性能各异的模型上验证了上述结论，并提出一种简单通用的即插即用方法，指导模型设计与优化。
原文摘要 · Abstract (English)
While deep learning is facing an homogenization across modalities led by Transformers, they are still challenged by shallow linear models in the time series forecasting task. Our hypothesis is that models should learn a direct link from past to future data points, which we identify as a learning dynamics capability. We develop an original $\texttt{PRO-DYN}$ nomenclature to analyze existing models through the lens of dynamics. Two observations thus emerge: $\textbf{1.}$ under-performing architectures learn dynamics at most partially, $\textbf{2.}$ the location of the dynamics block at the model end is of prime importance. Our systemic and empirical studies both confirm our observations on a set of performance-varying models with diverse backbones. We propose a simple plug-and-play methodology guiding model designs and improvements.
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。