It is used to solve two kinds of problems, function regression and stochastic time series prediction.Compared with BP and RBF neural network, PDNN has preferable approximation and robust performance.

• 通过与BP网络和RBF网络在函数拟合和混沌时间序列预测方面的对比实验,发现 PDNN具有非常良好的逼近性能和鲁棒性能。