投影寻踪学习网络及其在控制系统中的应用展望  

Projection Pursuit Learning Networks and Its Applications in Control Systems

作　　者：骆晨钟[1] 张志强[1] 邵惠鹤[1] 

机构地区：[1]上海交通大学自动化系

出　　处：《上海交通大学学报》1999年第5期631-634,共4页Journal of Shanghai Jiaotong University

摘　　要：介绍了投影寻踪学习网络（ＰＰＬＮ）的结构原理，给出了其基本学习算法．讨论了ＰＰＬＮ算法结构改进及优缺点，分析了它对多变量控制系统建模、软测量及控制器设计问题的适用性，并展望了ＰＰＬＮ良好的应用前景．Projection Pursuit Learning Network(PPLN) is a new type of feed forward neural networks emerged in recent years.Compared with other feed forward networks,PPLN is more parsimonious and learns more effectively.It is much suitable for high dimension problems with relatively small training sets.The mathematical principle of PPLN was explained and the basic training algorithms were given.Improvements and advantages of the network were discussed.Applicability of PPLN for identification,soft sensing and controller design problems were analyzed.It is shown that PPLN has potential value in practical multivariable control systems.

关 键 词：投影寻踪 学习网络 控制系统 多变量控制 PPLN 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TP273[自动化与计算机技术—控制科学与工程]