In this paper,in order to effectively speed up the weight convergence speed of neural network,a new back-propagation algorithm is proposed.

Based on the idea of standard back-propagation(BP) learning algorithm, an improved BP learning algorithm is presented.

Next, an Improved Back Propagation(IBP)algorithm is proposed, considering the drawbacks of the standard Back Propagation (BP) in the neural network theory.

Based on this framework, by making use of the back propagation method of computing partial derivatives, the maximization of mutual information can be achieved through the gradient ascent optimization process, and thus the discriminative training of HMM is performed.

In this study,the flood hydrograph at the forecasting section of tidal reach was considered as the superposition between flood waves from upstream and tide waves from downstream after bidirectional propagation.

A new combination method of multiple classifiers is presented and the method is controlled by a feed-forward neural net trained with the backpropagation technique.

As compared to the conventional BP algorithm, the first improvement of this algorithm is the minimization of a combined error criterion based on the sum of nonlinear and linear quadratic errors of the output neurons, the other is the using of a modified error backpropagation factor in the learning process.