Simulação do comportamento estocástico do algoritmo KLMS com diferentes kernels

The kernel least-mean-square (KLMS) algorithm is a popular algorithm in nonlinear adaptive filtering due to its simplicity and robustness. In kernel adaptive filtering, the statistics of the input to the linear filter depends on the kernel and its parameters. Moreover, practical implementations on systems estimation require a finite non-linearity model order. In order to obtain finite order models, many kernelized adaptive filters use a dictionary of kernel functions. Dictionary size also depends on the kernel and its parameters. Therefore, KLMS may have different performances on the estimation of a nonlinear system, the time of convergence, and the accuracy using a different kernel. In order to analyze the performance of KLMS with different kernels, this paper proposes the use of the Monte Carlo simulation of both steady-state and the transient behavior of the KLMS algorithm using different types of kernel functions and Gaussian inputs.