arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2210.01554stat.COcs.NAmath.NA
收藏

基于三次分层的高阶随机积分

Higher-order stochastic integration through cubic stratification

Nicolas Chopin, Mathieu Gerber

AI总结:

本研究提出两种依赖光滑性参数的无偏高阶随机积分估计量,结合三次分层与基于数值导数的控制变量法构建,分别实现最优误差和更低计算成本,中等求值次数下性能良好。

摘要:

我们针对函数$f$的积分$\u222b_{[0,1]^{s}}f(u) du$提出了两种全新的无偏估计量,其依赖于光滑性参数$r\in\mathbb{N}$。第一种估计量可精确积分次数$p<r$的多项式,当$f$具有$r$阶连续可微性时,能达到最优误差$n^{-1/2-r/s}$(其中$n$为$f$的求值次数)。第二种估计量计算成本更低,但仅适用于在$[0,1]^s$边界上取值为零的函数。两种估计量的构建结合了三次分层和基于数值导数的控制变量法。我们提供的数值证据表明,即使在$n$取中等值时,它们也表现出良好的性能。

英文摘要:

We propose two novel unbiased estimators of the integral $\int_{[0,1]^{s}}f(u) du$ for a function $f$, which depend on a smoothness parameter $r\in\mathbb{N}$. The first estimator integrates exactly the polynomials of degrees $p<r$ and achieves the optimal error $n^{-1/2-r/s}$ (where $n$ is the number of evaluations of $f$) when $f$ is $r$ times continuously differentiable. The second estimator is computationally cheaper but it is restricted to functions that vanish on the boundary of $[0,1]^s$. The construction of the two estimators relies on a combination of cubic stratification and control ariates based on numerical derivatives. We provide numerical evidence that they show good performance even for moderate values of $n$.

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。

↑