提出动态相关性生成方法，让合成时序数据更真实。
Dynamic Linear Coregionalization for Realistic Synthetic Multivariate Time Series
- 用动态线性协区化模型建模通道间时变相关性
- 在9个基准上微调后实现零样本预测提升
- 适合需要真实多变量时序数据的研究者
合成数据对时间序列基础模型（FMTS）训练至关重要，但现有生成器多假设相关性静态，缺乏真实的跨通道依赖。我们提出DynLMC，一种动态线性协区化模型，能捕捉时变、状态切换的相关性及跨通道滞后结构。该方法生成的合成多变量时间序列相关性动态与真实数据高度相似。在三个基础模型上使用DynLMC生成数据进行微调，均在九个基准上实现一致的零样本预测性能提升。结果表明，建模动态通道间相关性可显著增强FMTS的迁移能力，凸显数据驱动预训练的重要性。
原文摘要 · Abstract (English)
Synthetic data is essential for training foundation models for time series (FMTS), but most generators assume static correlations, and are typically missing realistic inter-channel dependencies. We introduce DynLMC, a Dynamic Linear Model of Coregionalization, that incorporates time-varying, regime-switching correlations and cross-channel lag structures. Our approach produces synthetic multivariate time series with correlation dynamics that closely resemble real data. Fine-tuning three foundational models on DynLMC-generated data yields consistent zero-shot forecasting improvements across nine benchmarks. Our results demonstrate that modeling dynamic inter-channel correlations enhances FMTS transferability, highlighting the importance of data-centric pretraining.
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。