ZeroDiff：基于先验信息扩散的零样本时间序列重建
ZeroDiff: Zero-Shot Time Series Reconstruction via Informed-Prior Diffusion
查看机构详情
- University of Pittsburgh（匹兹堡大学）
- Oak Ridge National Laboratory（橡树岭国家实验室）
机构由 AI 辅助整理，请以论文原文为准。
浏览论文内容
中文总结 AI 辅助
针对目标观测稀缺的零样本时间序列重建问题，提出ZeroDiff方法，利用外生变量构建先验并通过扩散校准误差，在真实数据集上显著优于现有方法。
中文摘要 AI 辅助
时间序列建模日益需要高质量的监督信号，然而目标观测数据仍然稀缺——外生输入广泛可得，但由于成本、基础设施或可访问性限制，目标测量值往往无法获取。能否利用在已观测位置训练的模型，重建从未采集过测量的目标时间序列？我们将此称为零样本时间序列重建。一种朴素的方法——直接将外生输入映射到目标——可以在未观测位置产生预测，但缺乏目标信号时，此类模型无法捕捉目标变量的内在动态，产生过度平滑的输出，低估极端值。这揭示了系统性误差，需要显式建模和校准。我们提出ZeroDiff，它仅从外生变量构建先验信息，然后通过扩散学习校准重建误差——在已观测位置训练，并泛化到未观测位置。在多个真实世界数据集上的实验表明，该方法显著优于现有方法。我们的代码可在该https URL获取。
英文摘要
Time series modeling increasingly demands high-quality supervision, yet target observations remain scarce - exogenous inputs are broadly available, but target measurements are often unavailable due to cost, infrastructure, or accessibility constraints. Can models trained on observed locations reconstruct target time series where measurements have never been collected? We term this zero-shot time series reconstruction. A naive approach - directly mapping exogenous inputs to targets - can yield predictions at unobserved locations, but without target signals, such models fail to capture the intrinsic dynamics of the target variable, producing overly smooth outputs that underestimate extremes. This reveals systematic errors that call for explicit modeling and calibration. We propose ZeroDiff, which constructs an informed prior from exogenous variables alone, then learns to calibrate reconstruction errors through diffusion - training on observed locations and generalizing to unobserved ones. Experiments across diverse real-world datasets demonstrate significant improvements over existing approaches. Our code is available at https://github.com/YingdaFan/ZeroDiff-ICML2026.
