Chebyshev Individual Adaptive Exponential Functional Link Filter for Nonlinear System Identification

Recently introduced adaptive exponential functional link networks (AEFLNs) are widely employed nonlinear filters with linear-in-the-parameters. In addition, sinusoidal basis functions are used with individually varying adaptive exponential factors for enhanced nonlinear system modelling. To improve the performance of these nonlinear filters, a new Chebyshev individual adaptive exponential functional link network (CIAEFLN) is designed in this paper. An independently varying exponential term is included in the Chebyshev polynomial as a nonlinear expansion parameter to improve the accuracy and convergence rate. Bounds on the learning rate and weight update have also been derived for the introduced CIAEFLN technique. Simulation outputs demonstrate the improved performance of the proposed algorithm as compared to the present techniques for various nonlinear system identification.