Increasing the initial convergence of distributed diffusion LMS algorithm by a new variable tap-length variable step-size method

Among strategies using in distributed adaptive networks diffusion based algorithms despite their scalability, robustness and steady state performance suffer from slow initial convergence. We propose a method to speed up this convergence rate by arranging the network nodes into subgroups, partitioning the tap weight vector and taking advantage of the larger step-size allowed for short filters. As our simulation results show, the proposed algorithm has a faster convergence rate as compared with conventional diffusion LMS algorithm and other algorithms have intended to increase the initial convergence rate of diffusion algorithms.