Parameter Identification of Finite-field Networks

This paper investigates the parameter identification problem of finite-field networks (FFNs). Firstly, an algebraic expression of FFNs is obtained by the technique of semi-tensor product, and then, the identification issue is converted into identifying the parameter matrices of linear system. Secondly, the forward and backward probability and their properties are addressed to facilitate the parameter identification procedure. Thirdly, the expectation maximization (EM) algorithm is proposed as an elaborate technique for parameter identification problem in FFNs. In addition, an useful index is introduced to describe the performance of network parameter identification. Finally, a typical example is employed to demonstrate the reliability of the proposed parameter identification approach.