改进的BP神经网络在流域产沙量预测中的应用

误差逆传播算法是多层前向网络的典型算法，但是其平方误差函数超曲面存在许多局部极小值，于是给出了基于输出空间的全局优化BP算法（global optimization back propagation algorithm, 简称GOBPA），应用GOBPA，建立黄河某流域年均产沙量的预测模型，结果表明，用GOBPA训练的多层前向神经网络能够以很高精度预报年均产沙量。

The Application of Improved BP Network in Amounts of Sediment Forecasting in a Watershed

Error back propagation algorithrm is the classical algorithm of multilayer forward net work, but many local minimum value lie on its hypersurface of the square error function, a global optimization algoritlim based on output space is proposed. Using the improved BP neural network,the prediction model of the amounts of a part of Yellow River is constructed .The results showed that the network can forecast the amounts of sediment precisely.