Application research of improved BP algorithm in process neural networks

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 Improved 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 applied to the forecast of aircraft EGT Index which proved the effectiveness of the algorithm as well as indicated the good future of PNN.