基于模糊神经网络的机场噪声烦恼度模型学习  

Learning method of airport noise annoyance model based on fuzzy neural network

作　　者：冯霞[1,2] 张聪颖[1] 卢敏[1,2] 

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300 [2]中国民航信息技术科研基地,天津300300

出　　处：《计算机工程与设计》2015年第12期3396-3401,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61139002);国家863高技术研究发展计划基金项目(2012AA063301);中国民用航空局科技基金项目(MHRD201130);中央高校科研业务经费基金项目(3122013P013;3122013C005);国家科技支撑计划课题基金项目(2014BAJ04B02)

摘　　要：鉴于机场噪声烦恼度模型的模糊规则量较大,采用传统的基于梯度的模糊神经学习算法存在计算量大、收敛速度慢、学习效率低的问题,提出一种基于模糊神经网络的机场噪声烦恼度模型混合学习方法。基于聚类思想,重新对模糊集合进行分组,采用先粗学习后细学习的间接学习方法;改进传统的基于梯度的模糊神经学习算法,将该算法应用到间接学习过程中,即混合学习方法。实验结果表明,该混合学习方法可以快速收敛,缩短学习时间,减少误差求解过程中的计算量,提高模型的学习效率。In view of the large fuzzy rules of airport noise annoyance model and the existing problems which are large amount of calculation, slow convergence speed, and low learning efficiency when using the conventional neural-fuzzy learning algorithm based on gradient, a mixed learning method of airport noise annoyance model based on fuzzy neural network was proposed. Based on the idea of clustering, the fuzzy sets were regrouped and the indirect learning method that learning from roughness to fineness was adopted. The traditional neural-fuzzy learning algorithm based on gradient was improved, and it was applied to the indirect learning method. The results show that the mixed method can improve the convergence speed, greatly shorten the learning time, reduce the amount of calculation in the process of solving the error, and improve the efficiency of the model learning.

关 键 词：机场噪声烦恼度模型 模型学习 模糊神经网络 改进的学习算法 混合学习方法 

分 类 号：TP39[自动化与计算机技术—计算机应用技术]