Stochastic behavior analysis of the Gaussian KLMS algorithm for a correlated input signal

Abstract

Abstract This paper derives a new model for the stochastic behavior of the Gaussian KLMS algorithm. The analysis considers the possibility of time correlated input vectors, a situation that cannot be modeled by existing models. Recursions are derived which predict both the transient and the steady-state behaviors of the algorithm for a time-varying dictionary. The model predictions show excellent agreement with Monte Carlo simulations in both modes of operation, providing significant improvement when compared to the accuracy of existing models.