arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2308.08035math.NAcs.NAstat.CO
收藏

置乱 Halton 点的增益系数

Gain coefficients for scrambled Halton points

Art B. Owen, Zexin Pan

AI总结:

本文研究置乱 Halton 点在随机化拟蒙特卡洛中的方差增益,证明其增益系数具有 $O(\log d)$ 的显式上界,并给出 $6\leqslant d\leqslant10^6$ 时的具体界。

摘要:

随机化拟蒙特卡洛通过对数字网络进行某些置乱,可对 $\int_{[0,1]^d}f(\boldsymbol{x})\\,\mathrm{d}\boldsymbol{x}$ 产生无偏估计;对任意 $f\in L^2[0,1]^d$,其方差为 $o(1/n)$。它还满足若干非渐近界,即方差不超过普通蒙特卡洛方差的某个 $\Gamma<\infty$ 倍。对于置乱 Sobol' 点,该量 $\Gamma$ 随 $d$ 指数增长。对于置乱 Faure 点,在任意维数下均有 $\Gamma\leqslant\exp(1)\doteq2.718$,但这些点在 $d$ 较大时使用不便。本文表明,Halton 序列的某些置乱所具有的增益低于一个显式界;当 $d\to\infty$ 时,该界为 $O(\log d)$,但对任意 $\varepsilon>0$ 都不是 $O((\log d)^{1-\varepsilon})$。对于 $6\leqslant d\leqslant10^6$,增益系数的上界始终不大于 $3/2+\log(d/2)$。

英文摘要:

Randomized quasi-Monte Carlo, via certain scramblings of digital nets, produces unbiased estimates of $\int_{[0,1]^d}f(\boldsymbol{x})\,\mathrm{d}\boldsymbol{x}$ with a variance that is $o(1/n)$ for any $f\in L^2[0,1]^d$. It also satisfies some non-asymptotic bounds where the variance is no larger than some $Γ<\infty$ times the ordinary Monte Carlo variance. For scrambled Sobol' points, this quantity $Γ$ grows exponentially in $d$. For scrambled Faure points, $Γ\leqslant \exp(1)\doteq 2.718$ in any dimension, but those points are awkward to use for large $d$. This paper shows that certain scramblings of Halton sequences have gains below an explicit bound that is $O(\log d)$ but not $O( (\log d)^{1-ε})$ for any $ε>0$ as $d\to\infty$. For $6\leqslant d\leqslant 10^6$, the upper bound on the gain coefficient is never larger than $3/2+\log(d/2)$.

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。

↑