English

An Accelerated Randomized Kaczmarz Algorithm

Numerical Analysis 2014-06-10 v2 Optimization and Control

Abstract

The randomized Kaczmarz (\RK\RK) algorithm is a simple but powerful approach for solving consistent linear systems Ax=bAx=b. This paper proposes an accelerated randomized Kaczmarz (\ARK\ARK) algorithm with better convergence than the standard \RK\RK algorithm on ill conditioned problems. The per-iteration cost of \RK\RK and \ARK\ARK are similar if AA is dense, but \RK\RK is much more able to exploit sparsity in AA than is \ARK\ARK. To deal with the sparse case, an efficient implementation for \ARK\ARK, called \SARK\SARK, is proposed. A comparison of convergence rates and average per-iteration complexities among \RK\RK, \ARK\ARK, and \SARK\SARK is given, taking into account different levels of sparseness and conditioning. Comparisons with the leading deterministic algorithm --- conjugate gradient applied to the normal equations --- are also given. Finally, the analysis is validated via computational testing.

Keywords

Cite

@article{arxiv.1310.2887,
  title  = {An Accelerated Randomized Kaczmarz Algorithm},
  author = {Ji Liu and Stephen J. Wright},
  journal= {arXiv preprint arXiv:1310.2887},
  year   = {2014}
}
R2 v1 2026-06-22T01:44:23.136Z