A stochastic Newton algorithm with data-adaptive step size

A stochastic Newton algorithm for adaptive filtering which is a generalization of the so-called normalized least mean-square algorithm is proposed. Expressions for the expected value of the algorithm weight error vector and second moment of the weight error-vector norm are derived, and have been verified by computer simulation. Experimental results which demonstrate the improved convergence of the algorithm over the algorithm after a large number of iterations are given.<
>