改进BP算法在过程神经网络中的应用

Application research of improved BP algorithm in process neural networks

过程神经网络是一种基于过程神经元的新型神经网络,其输入及权值皆为时序函数.针对基本BP算法的一些不足,将一种改进BP算法应用于前馈过程神经网络的训练之中,提高了网络的学习效率,扩展了过程神经网络的训练算法理论.在改进BP算法与网络训练的结合过程中,权函数及输入函数皆被用同一正交基函数展开.最后基于改进BP算法将过程神经网络应用于了飞机发动机尾气温度指数的预测当中,验证了算法的有效性,同时也展示了过程神经网络广泛的应用前景.

Process Neural Networks（PNN） is a kind of new ANN based on Process Neuron, its inputs and weights are all functions associated with ＇ time＇ In this paper, aiming at the disadvantages of BP algorithm, a kind of Improved BP algorithm was applied to the training of PNN which advanced the learning efficiency of PNN as well as expanded the algorithm theory of PNN to some extent. In the process of application of hnproved BP algorithm in PNN, the input functions and the network weight functions are represented as expansion of a same orthogonal function basis. At the end of the paper, on the basis of Improved BP algorithm, PNN was ap- plied to the forecast of aircraft EGT Index which proved the effectiveness of the algorithm as well as indicated the good future of PNN.