Recurrent Neural Network for the Identification of Nonlinear Dynamical Systems: A Comparative Study
摘要
This study compares three different recurrent neural network topologies in order to assess their estimated capabilities. The three neural networks under detailed examination are the Elman recurrent neural network (ERNN), the diagonal recurrent neural network (DRNN), and the Jordan recurrent neural network (JRNN). A dynamical backpropagation algorithm is applied for developing and updating the parameters linked to each of these neural nets. The comparative analysis is performed by considering one nonlinear dynamical system with varying degrees of complexity.