Upper and lower bounds of the misadjustment in the LMS algorithm

Author

Jaggi, S. ; Martinez, A.B.

Author_Institution

Dept. of Electr. Eng., Tulane Univ., New Orleans, LA, USA

Volume

38

Issue

1

fYear

1990

fDate

1/1/1990 12:00:00 AM

Firstpage

164

Lastpage

166

Abstract

The misadjustment is a dimensionless measure of the difference between the optimal Weiner performance and the actual performance of the LMS (least-mean-squares) algorithm. This difference is attributed to the gradient noise in the adaptive process. Present estimates relate the misadjustment to the convergence gain and the trace of the autocorrelation matrix. The authors establish tighter lower bounds on the estimate of this misadjustment. An upper bound on the misadjustment as a function of the convergence gain is found

Keywords

convergence of numerical methods; correlation methods; signal processing; LMS algorithm; actual performance; adaptive process; autocorrelation matrix; convergence gain; dimensionless measure; gradient noise; least-mean-squares; lower bounds; misadjustment; optimal Weiner performance; signal processing; upper bound; Algorithm design and analysis; Autocorrelation; Convergence; Delay estimation; Eigenvalues and eigenfunctions; Least squares approximation; Mean square error methods; Signal processing algorithms; Steady-state; Upper bound;

fLanguage

English

Journal_Title

Acoustics, Speech and Signal Processing, IEEE Transactions on

Publisher

ieee

ISSN

0096-3518

Type

jour

DOI

10.1109/29.45631

Filename

45631