Aplicacao Da Rede De Hopfield Na Estimacao Robusta De Regioes De Pertinencia Parametrica Para Modelos Lineares

This paper is concerned with the robust identification of linear model when the modeling error is assumed bounded. A modified Hopfield's Neural Network is developed to calculate a membership set for the model parameters. The valid-subspace technique is applied to obtain the internal parameters of the Hopfield's Neural Network. These parameters are explicitly computed to assure the network convergence. In this case, the equilibrium point represents a solution to robust estimation problem with unknown-but-bounded error. A comparative analysis with other robust estimation approaches is carried out by simulation examples.7311812