Distributed diffusion bias-compensated LMS for node-specific networks

In this paper, we study the problem of node-specific parameter estimation(NSPE) over distributed multi-agent networks, whose nodes have noise-corrupted regressor vectors. When the classic diffusion least mean square(LMS) algorithm is used in this situation, it results biased estimates of the nodal objectives. Therefore, we propose an online bias-compensated method to remove the bias introduced on the diffusion LMS results. Moreover, we investigate performance analysis in the mean and mean-square sense. Furthermore, we provide numerical experiments to illustrate and compare the robustness of our method under various distributed strategies and different network topologies.