English

Improved Algorithms for Adaptive Compressed Sensing

Data Structures and Algorithms 2018-04-26 v1 Information Theory math.IT

Abstract

In the problem of adaptive compressed sensing, one wants to estimate an approximately kk-sparse vector xRnx\in\mathbb{R}^n from mm linear measurements A1x,A2x,,AmxA_1 x, A_2 x,\ldots, A_m x, where AiA_i can be chosen based on the outcomes A1x,,Ai1xA_1 x,\ldots, A_{i-1} x of previous measurements. The goal is to output a vector x^\hat{x} for which xx^pCmink-sparse xxxq\|x-\hat{x}\|_p \le C \cdot \min_{k\text{-sparse } x'} \|x-x'\|_q\, with probability at least 2/32/3, where C>0C > 0 is an approximation factor. Indyk, Price and Woodruff (FOCS'11) gave an algorithm for p=q=2p=q=2 for C=1+ϵC = 1+\epsilon with \Oh((k/ϵ)\loglog(n/k))\Oh((k/\epsilon) \loglog (n/k)) measurements and \Oh(log(k)\loglog(n))\Oh(\log^*(k) \loglog (n)) rounds of adaptivity. We first improve their bounds, obtaining a scheme with \Oh(k\loglog(n/k)+(k/ϵ)\loglog(1/ϵ))\Oh(k \cdot \loglog (n/k) +(k/\epsilon) \cdot \loglog(1/\epsilon)) measurements and \Oh(log(k)\loglog(n))\Oh(\log^*(k) \loglog (n)) rounds, as well as a scheme with \Oh((k/ϵ)\loglog(nlog(n/k)))\Oh((k/\epsilon) \cdot \loglog (n\log (n/k))) measurements and an optimal \Oh(\loglog(n))\Oh(\loglog (n)) rounds. We then provide novel adaptive compressed sensing schemes with improved bounds for (p,p)(p,p) for every 0<p<20 < p < 2. We show that the improvement from O(klog(n/k))O(k \log(n/k)) measurements to O(kloglog(n/k))O(k \log \log (n/k)) measurements in the adaptive setting can persist with a better ϵ\epsilon-dependence for other values of pp and qq. For example, when (p,q)=(1,1)(p,q) = (1,1), we obtain O(kϵloglognlog3(1ϵ))O(\frac{k}{\sqrt{\epsilon}} \cdot \log \log n \log^3 (\frac{1}{\epsilon})) measurements.

Keywords

Cite

@article{arxiv.1804.09673,
  title  = {Improved Algorithms for Adaptive Compressed Sensing},
  author = {Vasileios Nakos and Xiaofei Shi and David P. Woodruff and Hongyang Zhang},
  journal= {arXiv preprint arXiv:1804.09673},
  year   = {2018}
}

Comments

To appear in ICALP 2018