A direction set based algorithm for adaptive filtering

Abstract

A new algorithm is developed for adaptive filtering applications. This algorithm is based on a direction set method and has a computational complexity of O(N) for each update of the system. The method exploits the structure of the objective function and maintains a set of near-conjugate directions with respect to the Hessian. This algorithm has a rapid rate of convergence that is comparable with that of the well-known RLS method. The performance of the algorithm is illustrated with adaptive filtering applications.