Optimal Combination Weight for Sparse Diffusion Least-Mean-Square based on Consensus Propagation

This paper considers distributed adaptive signal processing for tracking an unknown sparse parameter vector in large-scale networks. We propose a sparsity-promoting diffusion least-mean-square algorithm based on consensus propagation, which is an average consensus algorithm using message passing techniques. The main contributions of the paper are optimizing coefficients in the algorithm in terms of the steady-state error to achieve better convergence and robustness, and presenting the adaptive implementation.