A Novel Adaptive Algorithm for Estimation of Sparse Parameters in Non-Gaussian Noise

The goal of this paper is to propose a novel diffusion adaptive algorithm for estimation of sparse system with non-Gaussian noise. The proposed adaptive algorithm uses zero-norm regularization for accelerating the speed of convergence in sparse conditions and it uses maximum correntropy criterion (MCC) to access lower steady-state error in the presence of non-Gaussian noise. In order to reach more reduction in both transient and steady-state error, a diffusion strategy of the proposed algorithm is also employed where each local adaptive filter solves its own optimization problem and then diffuses information to its neighbors. The superiority of the proposed algorithm for non-Gaussian noises and sparse conditions is shown via computer simulations.