Identification of non-parametric FIR non-linear systems with low-degree interactive terms

Abstract

In this paper, an interactive term identification approach is proposed for identification of non-parametric nonlinear systems. The idea is to make a high-dimensional nonlinear identification problem into a number of low-dimensional problems and thus to effectively combat the problem of the curse of dimensionality. Convergence results are established in the paper and numerical results support the theoretical analysis and demonstrate that the proposed approach is an attractive alternative to existing nonlinear identification methods.