An Alternative Solution to the Dynamically Regularized RLS Algorithm

Ahstract-The recursive least-squares (RLS) algorithm should be explicitly regularized to achieve a satisfactory performance when the signal-to-noise ratio is low. However, a direct implementation of the involved matrix inversion results in a high complexity. In this paper, we present a recursive approach to the matrix inversion of the dynamically regularized RLS algorithm by exploiting the special structure of the correlation matrix. The proposed method has a similar complexity to the standard RLS algorithm. Moreover, the new method provides an exact solution for a fixed regularization parameter, and it has a good accuracy even for a slowly time-varying regularization parameter. Simulation results confirm the effectiveness of the new method.