arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2306.09147cs.LGcs.AI
收藏

具有时间不规则性的多元时间序列概率学习

Probabilistic Learning of Multivariate Time Series with Temporal Irregularity

City University of Hong Kong(香港城市大学)

Yijun Li, Cheuk Hang Leung, Qi Wu

AI总结:

针对非均匀间隔、变量错位的多元时间序列,提出端到端概率框架,用动态条件连续归一化流和分解对数似然建模联合分布及依赖关系,实验证明其更优且适应性更强。

摘要:

多元时间序列的概率预测对于各类下游任务都至关重要。大多数现有方法依赖于序列在所有变量之间具有均匀间隔且相互对齐。然而,现实世界中的多元时间序列往往存在时间不规则性,包括非均匀间隔和变量错位,这给准确预测带来了重大挑战。为应对这些挑战,我们提出了一个端到端框架,在建模时间不规则性的同时,捕捉任意连续时间点上变量的联合分布。具体而言,我们引入动态条件连续归一化流,以非参数方式建模数据分布,从而适应现实数据集中常见的复杂、非高斯特征。随后,通过利用经过精心分解的对数似然目标,我们的方法能够高效捕捉时间依赖和横截面依赖。在一系列真实世界数据集上的大量实验表明,与现有方法相比,我们的方法具有优越性和适应性。

英文摘要:

Probabilistic forecasting of multivariate time series is essential for various downstream tasks. Most existing approaches rely on the sequences being uniformly spaced and aligned across all variables. However, real-world multivariate time series often suffer from temporal irregularities, including nonuniform intervals and misaligned variables, which pose significant challenges for accurate forecasting. To address these challenges, we propose an end-to-end framework that models temporal irregularities while capturing the joint distribution of variables at arbitrary continuous-time points. Specifically, we introduce a dynamic conditional continuous normalizing flow to model data distributions in a non-parametric manner, accommodating the complex, non-Gaussian characteristics commonly found in real-world datasets. Then, by leveraging a carefully factorized log-likelihood objective, our approach captures both temporal and cross-sectional dependencies efficiently. Extensive experiments on a range of real-world datasets demonstrate the superiority and adaptability of our method compared to existing approaches.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑