In the problem of adaptive compressed sensing, one wants to estimate an approximately k-sparse vector x∈Rn from m linear measurements A1x,A2x,…,Amx, where Ai can be chosen based on the outcomes A1x,…,Ai−1x of previous measurements. The goal is to output a vector x^ for which ∥x−x^∥p≤C⋅k-sparse x′min∥x−x′∥q with probability at least 2/3, where C>0 is an approximation factor. Indyk, Price and Woodruff (FOCS'11) gave an algorithm for p=q=2 for C=1+ϵ with \Oh((k/ϵ)\loglog(n/k)) measurements and \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)) measurements and \Oh(log∗(k)\loglog(n)) rounds, as well as a scheme with \Oh((k/ϵ)⋅\loglog(nlog(n/k))) measurements and an optimal \Oh(\loglog(n)) rounds. We then provide novel adaptive compressed sensing schemes with improved bounds for (p,p) for every 0<p<2. We show that the improvement from O(klog(n/k)) measurements to O(kloglog(n/k)) measurements in the adaptive setting can persist with a better ϵ-dependence for other values of p and q. For example, when (p,q)=(1,1), we obtain O(ϵk⋅loglognlog3(ϵ1)) measurements.
@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}
}