A reduced-communication diffusion LMS algorithm based on multi-hop

In this paper, we propose a reduced-communication diffusion LMS algorithm that each node receives the intermediate estimates from only a subset of neighbors. Neighbor nodes are selected by a time-varying dynamic parameter. Unlike conventional diffusion LMS, information from multi-hop neighbors are participated in estimation to improve the approximate accuracy of a global cost function. Simulation results illustrate that the proposed algorithm not only outperforms other related methods but also reduces the communication load.