arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2603.05271math.NAcs.NA
收藏

使用中位数格点算法进行周期函数的最坏情况 $L_p$-近似

Worst-case $L_p$-approximation of periodic functions using median lattice algorithms

Zhejiang University(浙江大学);The University of Tokyo(东京大学);RICAM, Austrian Academy of Sciences(奥地利科学院计算与应用数学研究所)

Zexin Pan, Mou Cai, Josef Dick, Takashi Goda, Peter Kritzer

AI总结:

本文提出了一种基于中位数格点算法的周期函数最坏情况 $L_p$-近似方法,证明了在加权Korobov空间中高概率误差界,并展示了其在不同 $p$ 值下的最优近似性能。

摘要:

我们研究了从加权Korobov空间 $H_{d,α,γ}$ 中多变量周期函数在 $L_p([0,1]^d)$ 范数下的最坏情况近似,其中 $α>1/2$。我们分析了一种 \emph{中位数格点算法},该算法通过在超双曲线型索引集上使用 $R$ 个秩-1 格点采样规则,对系数进行近似,然后通过逐元素中位数聚合得到的系数估计器。对于奇数次重复 $R>1$ 和奇素数格点大小 $N$,我们证明了在 $L_\infty$ 和 $L_2$ 中的高概率误差界。插值后,结果适用于所有 $1 \le p\le\infty$。特别是,算法以高概率满足 \[ \mathrm{err}(H_{d,α,γ},L_p,A)\ \le\ C_{d,α,β,\boldsymbolγ,p}\, N^{- α+ (\frac12 - \frac1p)_+ + β}, \qquad 1 \le p\le\infty,\ β>0, \] 其中 $(x)_+ = \max\{x, 0\}$,$N$ 是函数评估次数,权重 $\boldsymbolγ$ 和常数 $C_{d,α,β,\boldsymbolγ,p}$ 与 $N$ 无关。对于 $p=\infty$,在可和条件 $\sum_{j=1}^\infty γ_j^{1/(2α)}<\infty$ 下,$C_{d,α,β,\boldsymbolγ,\infty}$ 与维度无关。这些结果扩展了最近关于 $L_2$ 中基于中位数的格点近似的分析,并补充了相关的多移位格点方法,表明中位数聚合在加权Korobov空间中实现了几乎最优的 $L_p$-近似速率(除对数因子和任意小的损失外)。

英文摘要:

We study the worst-case approximation of multivariate periodic functions from the weighted Korobov space $H_{d,α,\boldsymbolγ}$ with smoothness $α>1/2$ and weights $\boldsymbolγ$ in the Lebesgue norm $L_p([0,1]^d)$ for $1\le p\le\infty$. We analyze a \emph{median lattice algorithm} that reconstructs a truncated Fourier series by approximating the coefficients on a hyperbolic-cross-type index set using $R$ rank-1 lattice sampling rules with independent randomly chosen generating vectors, and then aggregating the resulting coefficient estimators via the componentwise median. For an odd number of repetitions $R$ and an odd prime lattice size $N$, where $R$ depends logarithmically on $N$, we prove high-probability error bounds in both $L_\infty$ and $L_2$. Interpolation then yields the result for all $ p \in [1,\infty]$. In particular, with high probability, the worst-case $L_p$-error satisfies, for any $β>0$, an upper bound of the following form \[ \ C_{d,α,β,\boldsymbolγ,τ,p}\,R^{\frac{1}{2}+(\frac12 - \frac1p)_+} N^{- α+ (\frac12 - \frac1p)_+ + β}, \] where $(x)_+ = \max\{x, 0\}$, and the constant $C_{d,α,β,\boldsymbolγ,τ,p}$ is independent of $N$ and $R$. For $p=\infty$, $C_{d,α,β,\boldsymbolγ,τ,\infty}$ is dimension-independent under the summability condition $\sum_{j=1}^\infty γ_j^{1/(2α)}<\infty$. These results extend recent analyses of median-based lattice approximation in $L_2$ and complement related multiple-shift lattice approaches, showing that median aggregation yields nearly optimal $L_p$-approximation rates (up to logarithmic factors and an arbitrarily small loss) in weighted Korobov spaces.

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。

↑