A New Modified Elman Neural Network with Stable Learning Algorithms for Identification of Nonlinear Systems

摘要

In this paper a new dynamic neural network structure based on the Elman Neural Network (ENN), for identification of nonlinear systems is introduced. The proposed structure has feedbacks from the outputs to the inputs and at the same time there are some connections from the hidden layer to the output layer, so that it is called as Output to Input Feedback, Hidden to Output Elman Neural Network (OIFHO ENN). The capability of the proposed structure for representing nonlinear systems is shown analytically. Stability of the learning algorithms is analyzed and shown. Encouraging simulation results reveal that the idea of using the proposed structure for identification of nonlinear systems is feasible and very appealing.