Randomized Kaczmarz Algorithm for Inconsistent Linear Systems: An Exact MSE Analysis

TL;DR

This paper provides an exact mean squared error analysis of the randomized Kaczmarz algorithm for inconsistent linear systems, offering precise formulas and demonstrating that previous bounds can significantly overestimate the error.

Contribution

It derives an exact MSE formula for RKA in inconsistent systems and shows how to average over noise distributions, improving understanding of the algorithm's performance.

Findings

Exact MSE formula for inconsistent systems

Previous bounds may be overly conservative

Numerical simulations confirm accuracy of formulas

Abstract

We provide a complete characterization of the randomized Kaczmarz algorithm (RKA) for inconsistent linear systems. The Kaczmarz algorithm, known in some fields as the algebraic reconstruction technique, is a classical method for solving large-scale overdetermined linear systems through a sequence of projection operators; the randomized Kaczmarz algorithm is a recent proposal by Strohmer and Vershynin to randomize the sequence of projections in order to guarantee exponential convergence (in mean square) to the solutions. A flurry of work followed this development, with renewed interest in the algorithm, its extensions, and various bounds on their performance. Earlier, we studied the special case of consistent linear systems and provided an exact formula for the mean squared error (MSE) in the value reconstructed by RKA, as well as a simple way to compute the exact decay rate of the error. In this work, we consider the case of inconsistent linear systems, which is a more relevant scenario for most applications. First, by using a "lifting trick", we derive an exact formula for the MSE given a fixed noise vector added to the measurements. Then we show how to average over the noise when it is drawn from a distribution with known first and second-order statistics. Finally, we demonstrate the accuracy of our exact MSE formulas through numerical simulations, which also illustrate that previous upper bounds in the literature may be several orders of magnitude too high.