有理式二阶前馈型神经网络

文章提出了二阶有理式多层前馈神经网络的数学模型。有理式多层神经网络的思想来源于函数逼近理论中的有理式逼近。有理式前馈神经网络模型是传统前俯神经网络模型的推广，能有效地求解函数逼近问题。文章给出了有理式多层神经网络的学习算法，即误差反传播学习算法。就计算复杂度而言，有理式神经网络的学习算法与传统的多层神经网络反传播算法是同阶的。文章还给出了函数逼近和模式识别两个应用实例，实验结果说明二阶有理式多层神经网络在解决传统的问题上是有效的。

Two order rational fraction feedforward neural networks

This paper deals with the mathematic model of two order rational fraction feedforward neural networks. The idea of rational fraction feedforward neural networks comes from rational fraction approximation in function approximation theories, This model is the general form of traditional multiplayer neural networks, and it is adaptive to solve function approximation issues. It proposes the learning paradigm for rational fraction multiplayer neural networks, i.e, error back propagation algorithm. From the view of computing colnplex, the learning algorithm for rational fraction multiplayer neural networks is the same degree as traditional multiplayer neural networks. This paper also proposes two application examples, and one is function approximation issue and the other is a pattern recognition problem. Experiment results illustrate the fitness of two order rational fraction multiolaver neural networks in solving traditional oroblems.