Diagonal recurrent neural network with output feedback and its application

In this paper, a new diagonal recurrent neural network with output feedback model was proposed and applied to ship rolling time series prediction. On the basis of diagonal recurrent neural network, the output was feedback to hidden layer to form a new neural network structure. The new neural network and its mathematical model were given at first. Second, the learning algorithm and the update rule were presented in detail. Because of internal self-feedback, the diagonal recurrent neural network with output feedback can learn nonlinear dynamic system without knowing system order. Finally, the simulation results proved that the proposed model has good online prediction capability.