This paper introduces a fast adaptive polynomial filtering algorithm,
called LS-LMS algorithm, and analyzes its connections with RLS and wit
h several QR decomposition based adaptive algorithms introduced in (Li
u, 1995) and (Niemisto et al., 1996). Since the time-shift invariance
property of the input data (Haykin, 1996, p. 763) is not required for
the input vector, the algorithm is well suited for the identification
of polynomial models. A noise cancelation application exemplifies the
benefits of using the new algorithm. (C) 1998 Published by Elsevier Sc
ience B.V. All rights reserved.