Nonparametric Tracking for Time-Varying Nonlinearities Using the Kernel Method

In the real world, there is a number of systems that are nonlinear and time-varying. Modelling and nonparametric tracking of time-varying nonlinear characteristics are famous and playing important role in system identification. The weighted least-squares method has been used for the identification of such systems. The main objective of this paper is to track the time-varying nonlinearities using the kernel estimation method and balance between variance and bias. The optimal solution of the bias-variance trade-off has been evaluated by the zero-one forgetting weighted kernel method. The simulation results indicate that by tuning the bandwith and time horizont, the mean square error (MSE) can be reduced.