A Kernel Recursive Mixed Error Criterion Algorithm for Chaotic Time Series Prediction

In this paper, a new adaptive filtering algorithm with kernel method, named as kernel recursive mixed-error criterion (KRMEC), is derived in the context of nonlinear signal processing, which uses two-order error and correntropy schemes to build a novel mixed error cost-function. Then, we kernelized it to get KRMEC. The proposed KRMEC algorithm is used for chaotic time series prediction (CTST) under impulsive-noise environment. To evaluate the KRMEC algorithm, several simulations are setup to get the superiority of the proposed algorithm.