Looping LMS versus fast least squares algorithms: who gets there first?

Abstract

This paper analytically compares, in terms of the convergence time, fast least squares estimation algorithms for channel identification and equalization to looping LMS (LLMS), a scheme which repeatedly applies the least mean squares algorithm to a block of received data. In this study, the convergence time is defined as the actual time (in seconds) taken by an algorithm to reach a desired performance. The old theme on LMS and fast least squares algorithms convergence is revisited from a novel perspective: the comparison is made from a complexity viewpoint, which not only takes into account the statistical properties of studied algorithms but also the number of floating point operations.