PROJECT TITLE:

Generalized Correntropy for Robust Adaptive Filtering - 2016

ABSTRACT:

As a robust nonlinear similarity measure in kernel house, correntropy has received increasing attention in domains of machine learning and signal processing. In particular, the most correntropy criterion (MCC) has recently been successfully applied in strong regression and filtering. The default kernel perform in correntropy is the Gaussian kernel, which is, of course, not invariably the best alternative. In this paper, we tend to propose a generalized correntropy that adopts the generalized Gaussian density (GGD) operate as the kernel, and gift some important properties. We tend to additional propose the generalized maximum correntropy criterion (GMCC) and apply it to adaptive filtering. An adaptive algorithm, referred to as the GMCC algorithm, is derived, and the steadiness downside and steady-state performance are studied. We tend to show that the proposed algorithm is terribly stable and can achieve zero chance of divergence (POD). Simulation results ensure the theoretical expectations and demonstrate the fascinating performance of the new algorithm.