Operation property | The LMS-based algorithm minimizes the weighted sum of the squared error for stochastic signals. The speed of convergence of the LMS-based algorithm depends on the step-size and the eigenvalue spread of R_{y}(0 < μ < 2/λ_{max}, λ_{max} denotes the largest eigenvalue of R_{y}) | The RLS-based algorithm minimizes the weighted sum of the squared error for deterministic signals. The speed of convergence of the RLS-based algorithm does not depend on the condition number of R_{y}, which leads to, in general, an order of magnitude faster than that of the LMS-based algorithm. |