Diffusion Generalized Kernel Maximum Correntropy Criterion Algorithm With Variable Center: Establishment and Performance Analysis

The existing adaptive filtering algorithms in distributed networks do not take into account the case of non-zero mean non-Gaussian noise. In response to this issue, a robust diffusion kernel adaptive filtering algorithm, which is called the diffusion generalized kernel maximum correntropy criterion with variable center (DGKMCC-VC) algorithm, is proposed. The algorithm aims to address a noisy environment with non-zero mean by dynamically adjusting the center of the kernel function in the generalized correntropy. An online vector quantization method is also introduced, which can be used to reduce the complexity of the algorithm. In addition, the convergence condition and steady-state performance of the algorithm are analyzed. Finally, simulation results verify the robustness and superiority of the algorithm in dealing with non-Gaussian noise environments with non-zero mean in distributed models.