Much Faster Algorithms for Matrix Scaling
Abstract
We develop several efficient algorithms for the classical \emph{Matrix Scaling} problem, which is used in many diverse areas, from preconditioning linear systems to approximation of the permanent. On an input matrix , this problem asks to find diagonal (scaling) matrices and (if they exist), so that -approximates a doubly stochastic, or more generally a matrix with prescribed row and column sums. We address the general scaling problem as well as some important special cases. In particular, if has nonzero entries, and if there exist and with polynomially large entries such that is doubly stochastic, then we can solve the problem in total complexity . This greatly improves on the best known previous results, which were either or . Our algorithms are based on tailor-made first and second order techniques, combined with other recent advances in continuous optimization, which may be of independent interest for solving similar problems.
Cite
@article{arxiv.1704.02315,
title = {Much Faster Algorithms for Matrix Scaling},
author = {Zeyuan Allen-Zhu and Yuanzhi Li and Rafael Oliveira and Avi Wigderson},
journal= {arXiv preprint arXiv:1704.02315},
year = {2017}
}