多输入Laguerre正交多项式前向神经网络权值与结构确定法

Weights and Structure Determination of Multi-input Laguerre-orthogonal-polynomial Feed-forward Neural Network

为克服BP神经网络模型及其学习算法中的固有缺陷,根据多项式插值和逼近理论,构造出一种以Laguerre正交多项式作为隐层神经元激励函数的多输入前向神经网络模型。针对该网络模型,提出了权值与结构确定法,以便快速、自动地确定该网络的最优权值和最优结构。计算机仿真与实验结果显示:该算法是有效的,并且通过该算法所得到的网络具有较优的逼近性能和良好的去噪能力。

In order to remedy the inherent weaknesses of the back-propagation（BP） neural-network model and its learning algorithm,a multi-input Laguerre-orthogonal-polynomial feed-forward neural network（MILOPNN） was constructed,which is based on the theory of polynomial interpolation and approximation.Then,a new kind of weights-and-structure-determination（WASD） algorithm was proposed to determine the optimal weights and structure of the MILOPNN quickly and automatically.Computer simulation and experiment results further substantiate the efficacy of the WASD algorithm,as well as the relatively good abilities of approximation and denoising of the MILOPNN model equipped with the WASD algorithm.