Approximate Sparse Recovery: Optimizing Time and Measurements
Abstract
An approximate sparse recovery system consists of parameters , an -by- measurement matrix, , and a decoding algorithm, . Given a vector, , the system approximates by , which must satisfy , where denotes the optimal -term approximation to . For each vector , the system must succeed with probability at least 3/4. Among the goals in designing such systems are minimizing the number of measurements and the runtime of the decoding algorithm, . In this paper, we give a system with measurements--matching a lower bound, up to a constant factor--and decoding time , matching a lower bound up to factors. We also consider the encode time (i.e., the time to multiply by ), the time to update measurements (i.e., the time to multiply by a 1-sparse ), and the robustness and stability of the algorithm (adding noise before and after the measurements). Our encode and update times are optimal up to factors.
Cite
@article{arxiv.0912.0229,
title = {Approximate Sparse Recovery: Optimizing Time and Measurements},
author = {Anna C. Gilbert and Yi Li and Ely Porat and Martin J. Strauss},
journal= {arXiv preprint arXiv:0912.0229},
year = {2014}
}