Application of an ANNs with sensitive ability to hydrologic forecast

A new strategy involved in training sample and algorithm for networks weights is put forward in this paper in order to fully seek information of original data and expedite convergence rate.A forgetting factor and an expecting factor varying with time are designed into training.On the other hand,an exponential energy function is introduced for forming a new algorithm to modify the weights of ANNs.Two case studies are presented.The results show that the suggested strategy is reasonable and functional.