Robust training algorithm of multilayered neural networks for identification of nonlinear dynamic systems

Linear isotropic dielectrics constitute an important class of linear systems. The paper proposes a novel approach for assigning a suitable structure to describe the dependence of the dynamic behaviour of linear dielectrics on an external parameter of influence. A signal flow graph approach is adopted to arrive at the most general system structure and its subcases. A complex plane representation is used to establish a relationship between the system structure and the pattern of the constant-frequency variable-parameter plots. The one-to-one correspondence between loci patterns and system structure provides a rational basis for the selection of a suitable system structure.