Parallel recursive least squares algorithms for adaptive Volterra filters

Abstract

The authors present a parallel but approximate version of the exact recursive least squares algorithm. This algorithm, called the parallel recursive least squares (PRLS) algorithm, has been applied to adaptive Volterra filters. It has advantages of reduced cost per iteration and substantial reduction in computational time per iteration if more than one processor is used. The algorithm has the potential of providing a flexible tradeoff in terms of rate of convergence and computational cost.<>