Optimal Sketching for Kronecker Product Regression and Low Rank Approximation
Abstract
We study the Kronecker product regression problem, in which the design matrix is a Kronecker product of two or more matrices. Given for where for each , and , let . Then for , the goal is to find that approximately minimizes . Recently, Diao, Song, Sun, and Woodruff (AISTATS, 2018) gave an algorithm which is faster than forming the Kronecker product Specifically, for their running time is , where nnz is the number of non-zero entries in . Note that nnz can be as large as . For and , they achieve a worse bound of . In this work, we provide significantly faster algorithms. For , our running time is , which has no dependence on nnz. For , our running time is , which matches the prior best running time for . We also consider the related all-pairs regression problem, where given , we want to solve , where consist of all pairwise differences of the rows of . We give an time algorithm for , improving the time needed to form . Finally, we initiate the study of Kronecker product low rank and low -rank approximation. For input as above, we give time algorithms, which is much faster than computing .
Keywords
Cite
@article{arxiv.1909.13384,
title = {Optimal Sketching for Kronecker Product Regression and Low Rank Approximation},
author = {Huaian Diao and Rajesh Jayaram and Zhao Song and Wen Sun and David P. Woodruff},
journal= {arXiv preprint arXiv:1909.13384},
year = {2019}
}
Comments
A preliminary version of this paper appeared in NeurIPS 2019