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 2005.08926cs.LGstat.ML
收藏

不规则时间序列的神经控制微分方程

Neural Controlled Differential Equations for Irregular Time Series

Mathematical Institute, University of Oxford(牛津大学数学研究所);The Alan Turing Institute(艾伦·图灵研究所)

Patrick Kidger, James Morrill, James Foster, Terry Lyons

摘要:

神经常微分方程是一个有吸引力的选择建模时间动态。然而,一个根本的问题是,常微分方程的解是由其初始条件决定的,并且没有基于后续观测来调整轨迹的机制。在这里,我们演示了如何通过\emph{controlled differential equations}的良好理解的数学解决这一问题。由此产生的\emph{neural controlled differential equation}模型是直接适用于一般设置的部分观察到的不规则采样的多变量时间序列,和(不像以前的工作在这个问题上),它可以利用内存高效的伴随为基础的反向传播,即使在整个观察。我们证明了我们的模型在一系列数据集的实证研究中与类似(基于ODE或RNN的)模型相比具有最先进的性能。最后,我们提供的理论结果证明普遍的近似,我们的模型包含替代ODE模型。

英文摘要:

Neural ordinary differential equations are an attractive option for modelling temporal dynamics. However, a fundamental issue is that the solution to an ordinary differential equation is determined by its initial condition, and there is no mechanism for adjusting the trajectory based on subsequent observations. Here, we demonstrate how this may be resolved through the well-understood mathematics of \emph{controlled differential equations}. The resulting \emph{neural controlled differential equation} model is directly applicable to the general setting of partially-observed irregularly-sampled multivariate time series, and (unlike previous work on this problem) it may utilise memory-efficient adjoint-based backpropagation even across observations. We demonstrate that our model achieves state-of-the-art performance against similar (ODE or RNN based) models in empirical studies on a range of datasets. Finally we provide theoretical results demonstrating universal approximation, and that our model subsumes alternative ODE models.

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。

↑