Title :
The optimum scalar data nonlinearity in LMS adaptation for arbitrary IID inputs
Author :
Douglas, Scott C. ; Meng, Teresa H Y
Author_Institution :
Inf. Syst. Lab., Stanford Univ., CA, USA
fDate :
6/1/1992 12:00:00 AM
Abstract :
The authors show that the optimum nonlinear scale operation upon the elements of the observation vector in the LMS algorithm is exactly x/(1+μx2) for any independent stochastic data input and any noise density. Moreover, use of such a nonlinearity can yield a significant performance improvement in fast adaptation situations
Keywords :
least squares approximations; vectors; IID inputs; LMS algorithm; independent stochastic data input; noise density; observation vector; optimum nonlinear scale operation; optimum scalar data nonlinearity; Algorithm design and analysis; Convergence; Echo cancellers; Finite impulse response filter; Least squares approximation; Noise cancellation; Noise generators; Signal processing algorithms; Stochastic processes; Stochastic resonance;
Journal_Title :
Signal Processing, IEEE Transactions on
