A Kernel Affine Projection Maximum Correntropy Criterion Algorithm Based on the q-Rényi Kernel

This work, a new kernel affine projection algorithm called q-Rényi kernel-affine-projection-maximum correntropy-criterion (qKAPMCC) is proposed. The qKAPMCC algorithm is different from the traditional kernel affine projection algorithm: the cost-function for learning algorithm is derived by the correntropy-criterion, and the q-Rényi kernel function replaces the Gaussian kernel in RKHS. The qKAPCC have an outperformance in the NCE under the impulse-noise. Simulation in different non-Gaussian noise environments express high performance of the proposed qKAPMCC algorithm compared to other kernel adaptive filters.