基于Hessian矩阵的模糊优选BP算法及其应用  

Back propagation learning algorithm with Hessian matrix for fuzzy optimal neural networks

作　　者：刘金禄[1] 陈守煜[1] 

机构地区：[1]大连理工大学土木水利学院,辽宁大连116024

出　　处：《水利学报》2004年第4期33-37,共5页Journal of Hydraulic Engineering

摘　　要：在模糊优选神经网络理论的基础上,提出了等效误差函数的概念。建立了基于误差函数的Hessian矩阵的一种BP算法。该方法应用数值计算方法中的Newton迭代法显著提高了计算中的学习效率。将该方法应用于大连市经济、水资源和环境规划的决策分析中,取得了满意的效果。On the basis of theory for fuzzy optimal neural networks, the concept of equivalent error function is introduced to establish a new back propagation learning(BP)algorithm with Hessian matrix to promote the calculation.The Newton's iteration method is applied to promote the training efficiency and accelerate the convergence.This method is used to calculate the decision-making analysis for economy,water resources and environment planning of the Dalian City.

关 键 词：模糊优选 神经网络 误差函数 BP算法 

分 类 号：O159[理学—数学] TV213[理学—基础数学]