Reconstruction under outliers for Fourier-sparse functions

Reconstruction under outliers for Fourier-sparse functions
复制标题

傅里叶稀疏函数异常值下的重建

DOI:
10.1137/1.9781611975994.124
复制
发表时间:
2020
期刊:
Proceedings of the 2020 ACM-SIAM Symposium on Discrete Algorithms
影响因子:
--
通讯作者:
De, Anindya
De, Anindya
中科院分区:
--
文献类型:
--
作者:
Chen, Xue;De, Anindya

文献摘要

参考文献

被引文献

相似文献

我们考虑了在存在离群噪声的情况下学习具有稀疏傅立叶谱的未知数的问题。特别地,该算法可以访问(未知)f的噪声预言,使得(i)f的傅立叶谱是稀疏的;(ii)在任何查询点tx,预言返回y,使得概率为1-ρ,|y-f(x)|≤ ε。然而,对于概率p,误差f(x)可以是任意大的。我们研究了离散立方体{0,1}和环面[0,1)上的傅立叶稀疏函数,并且对于这两个域,我们设计了有效的算法,该算法可以容忍任何ρ <1/2的离群值。我们注意到,低次多项式的类似问题最近已经在几个作品[AK03,GZ 16,KKP 17]中进行了研究,并且在该设置中已知类似的算法保证。以致于|y-f(x)|> ε是随机分布的,我们还研究了异常值位于相反位置的情况。特别地,我们证明了在环面上,假设傅里叶变换满足一定的粒度条件,存在一个样本有效算法来容忍ρ = Ω(1)的离群值分数,并且进一步地,如果没有这样的粒度条件,这是不可能的。最后,虽然不是主要的推力,我们的技术还允许我们在存在对抗性离群噪声的情况下对超立方体上的低次函数的学习进行非平凡的改进。我们的技术联合收割机结合了压缩感知,稀疏傅立叶变换,链式参数和复杂分析等多种工具。
We consider the problem of learning an unknownfwith a sparse Fourier spectrum in the presence of outlier noise. In particular, the algorithm has access to a noisy oracle for (an unknown)fsuch that (i) the Fourier spectrum offisk-sparse; (ii) at any query pointx, the oracle returnsysuch that with probability 1 –ρ, |y–f(x)| ≤ε. However, with probability p, the errory–f(x) can be arbitrarily large.We study Fourier sparse functions over both the discrete cube {0, 1}nand the torus [0, 1) and for both these domains, we design efficient algorithms which can tolerate anyρ< 1/2 fraction of outliers. We note that the analogous problem for low-degree polynomials has recently been studied in several works [AK03, GZ16, KKP17] and similar algorithmic guarantees are known in that setting.While our main results pertain to the case where the location of the outliers, i.e.,xsuch that |y–f(x)| >εis randomly distributed, we also study the case where the outliers are adversarially located. In particular, we show that over the torus, assuming that the Fourier transform satisfies a certaingranularitycondition, there is a sample efficient algorithm to tolerateρ= Ω(1) fraction of outliers and further, that this is not possible without such a granularity condition. Finally, while not the principal thrust, our techniques also allow us non-trivially improve on learning low-degree functionsfon the hypercube in the presence of adversarial outlier noise.Our techniques combine a diverse array of tools from compressive sensing, sparse Fourier transform, chaining arguments and complex analysis.
DOI: 10.4230/oasics.sosa.2019.19
发表时间: 2018-09
期刊: --
影响因子: --
作者:
Sushrut Karmalkar;Eric Price
通讯作者: Sushrut Karmalkar;Eric Price
一种通过简单傅里叶变换重建信号的通用采样方法
DOI: 10.1145/3313276.3316363
发表时间: 2018
期刊: Proceedings of the 51st Annual ACM SIGACT Symposium on Theory of Computing
影响因子: --
作者:
I. Čižnár;A. Hoštacká;C. González;K. Krovacek
通讯作者: K. Krovacek
DOI: --
发表时间: 1997
期刊:
影响因子: --
作者:
P. Borwein;T. Erdélyi
通讯作者: T. Erdélyi
A·哈贝:(1989)
DOI: --
发表时间: --
期刊:
影响因子: --
作者:
通讯作者: --
将 L 1 的子空间嵌入到 l N 1 中
DOI: --
发表时间: 1990
期刊:
影响因子: --
作者:
M. Talagrand
通讯作者: M. Talagrand