Given data X∈Rn×d and labels y∈Rn the goal is find w∈Rd to minimize ∥Xw−y∥2. We give a polynomial algorithm that, \emph{oblivious to y}, throws out n/(d+n) data points and is a (1+d/n)-approximation to optimal in expectation. The motivation is tight approximation with reduced label complexity (number of labels revealed). We reduce label complexity by Ω(n). Open question: Can label complexity be reduced by Ω(n) with tight (1+d/n)-approximation?
@article{arxiv.2305.07486,
title = {Reduced Label Complexity For Tight $\ell_2$ Regression},
author = {Alex Gittens and Malik Magdon-Ismail},
journal= {arXiv preprint arXiv:2305.07486},
year = {2023}
}