Diffusion generalized maximum correntropy criterion algorithm for distributed estimation over multitask network

Adopting mean-square error (MSE) criterion, distributed estimation algorithms achieve desirable performance if the background noise is drawn from the Gaussian distribution. However, these algorithms may degrade considerably in non-Gaussian situations, especially for the impulsive-noise. The diffusion generalized maximum correntropy criterion (D-GMCC) algorithm is proposed in this article to address this problem. Furthermore, we contribute in multiple tasks problem, which differs from single-task estimation problem. The relatedness of tasks among nodes is also studied to uncover its impact on estimation performance. Simulations illustrate that the proposed algorithm achieves desirable performance and outperforms other related methods. The results on the mean and mean square stability analysis of the proposed algorithm are also provided.