z2-proportionate diffusion LMS algorithm with mean square performance analysis

Recently, a proportionate-type diffusion LMS algorithm has been proposed by minimizing the mean-square deviation of an intermediate estimate in sparse distributed estimation problems. This algorithm enhances the convergence speed regardless of the sparseness of the vector of interest, but the gain calculation demands huge computational resources. We develop a computationally efficient version of this algorithm that maintains the advantage of the conventional algorithm. We also analyze the mean-square performance of the proposed algorithm without assuming Gaussian distribution of unknown vector estimates (the usual assumption in analysis of proportionate adaptive filters).