Robust diffusion affine projection generalized maximum correntropy criterion algorithm over networks

Abstract

Generalized maximum correntropy criterion (GMCC) achieves desirable performance for system identification under non-Gaussian assumption. Combining diffusion strategy with affine projection generalized maximum correntropy criterion (APGMCC) algorithm, in this paper, we propose a new distribution type algorithm named diffusion APGMCC (DAPGMCC) algorithm, where every node in the diffusion network only exchanges information within its neighborhoods. Utilizing the Logarithmic function, we creat a new form of variable step-size DAPGMCC which splits the adaptation step into two steps (updation and combination). Simulation results demonstrate that the proposed two algorithms achieve good performance, compared with other related algorithms.