arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2401.09299math.PR
收藏

受控粗糙路径的尺度化二次变差与分数阶扩散过程的参数估计

Scaled quadratic variation for controlled rough paths and parameter estimation of fractional diffusions

James-Michael Leahy, Torstein Nilssen

AI总结:

本文提出有限γ-尺度化二次变差概念,证明受控路径继承该性质并满足Itô等距型公式,进而给出基于单条样本路径的γ一致估计量,并将其应用于分数阶扩散过程中Hurst指数与噪声向量场参数的估计,明确了收敛速率。

摘要:

我们引入了在给定区间上沿一列划分的有限γ-尺度化二次变差的概念。这一概念在历史上可追溯至Gladyshev(1961)以及Klein与Giné(1975)对高斯过程的研究,它涵盖了Hurst指数为H的分数布朗运动(fBM),后者具有有限的1-2H-尺度化二次变差。我们证明:在M. Gubinelli意义下由具有有限γ-尺度化二次变差的路径所控制的路径会继承这一性质,并且相应的尺度化二次变差满足一个Itô等距型公式。此外,我们证明了定量误差界,建立了受控路径与控路径的尺度化二次变差收敛速率之间的关系。另外,我们基于单条样本路径引入了参数γ的一致估计量,并给出了定量误差界。我们将这些结果应用于分数阶扩散过程的参数估计。我们的研究结果明确给出了Hurst指数以及噪声向量场中参数的估计收敛速率。论文最后以数值实验验证了我们的理论发现。

英文摘要:

We introduce the concept of finite $γ$-scaled quadratic variation along a sequence of partitions for paths on a given interval. This concept, with historical roots in the study of Gaussian processes by Gladyshev (1961) and Klein \& Giné (1975), includes the fractional Brownian motion (fBM) with Hurst index $H$, which has finite $1-2H$-scaled quadratic variation. We show that a path that is controlled by a path with finite $γ$-scaled quadratic variation in the sense of M. Gubinelli inherits this property, and the corresponding scaled quadratic variation satisfies an Itô-isometry type formula. Moreover, we prove quantitative error bounds that establish a relationship between the convergence rates of the scaled quadratic variation of the controlled path and that of the controlling path. Additionally, we introduce a consistent estimator for the parameter $γ$ based on a single sample path, complete with quantitative error bounds. We apply these results to the parameter estimation for fractional diffusions. Our findings specify convergence rates for the estimation of both the Hurst index and the parameters in the noise vector fields. The paper concludes with numerical experiments that substantiate our theoretical findings.

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。

↑