Nonlinear system identification using adaptive Chebyshev neural networks

Abstract

A new adaptive Chebyshev neural networks (ACNN) algorithm for the purpose of complex nonlinear system identification was proposed. In the proposed algorithm, the activation function of hidden units was defined by Chebyshev polynomials in the neural networks. The efficient algorithm for complex nonlinear system identification was constructed, which integrated Chebyshev neural networks with adaptive learning strategy to improve the identification accuracy and convergence rate. Furthermore, the networks algorithm was improved so that the applications becomed extensive. Then the ACNN directly learned dynamic characters of nonlinear system and identified it. The simulation results show that the ACNN algorithm have much less computation and high accuracy in the problem of complex nonlinear system identification.