A New on-Line Recursive Learning Algorithm for Recurrent Neural Network
一种回归神经网络的快速在线学习算法

In this paper a new on-line recursive learning algorithm for recurrent neural network is proposed. It overcomes the disadvantage of the slow convergence of the recurrent BP algorithm. The real-time learning ability and the fast convergence of the recurrent network model of nonlinear dynamical system have been obtained by introducing the forgetting factor in the objective function and the maximum likelihood estimation principle. Simulation results show that the proposed algorithm performs better than the traditional recurrent BP algorithm.