Asymptotically optimal smoothing of averaged LMS estimates for regression parameter tracking (Q1614388)

This paper considers the estimates formed by the LMS (least mean-squares) algorithm for a standard linear regression. It is shown that smoothing the LMS estimates using a matrix updating will lead to smoothed estimates with optimum tracking properties, also in case the true parameters are slowly changing as a random walk. It is also shown that the same accuracy can be obtained for a modified algorithm, SLAMS, which is based on averages and requires far fewer computations.