DHNN优化设计新方法及在板形模式识别的应用  

A novel method of optimal designing DHNN and applied to flatness pattern recognition

作　　者：张秀玲[1] 张志强[1] 

机构地区：[1]燕山大学电气工程学院,河北秦皇岛066004

出　　处：《智能系统学报》2008年第3期250-253,共4页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金资助项目(50374058);燕山大学博士基金资助项目(B70)

摘　　要：基于离散Hopfield神经网络(DHNN)的联想记忆能力,提出了随机扰动优化设计DHNN的新方法.该方法降低了DHNN对权值矩阵的苛刻要求,避免进入伪稳定点;并将其用于板形模式识别,采用勒让德多项式表示常见的6种板形基模式,不需大量的测试样本来训练网络,是一种更简单、实用的板形模式识别新方法,为实现板形控制提供依据,仿真结果证明了这种方法的可行性.A novel method of optimal designing DHNN is proposed based on random destabilization and its associative memory. This method reduces the harshness requirement of the weight matrix and avoids getting into the pseudo stability point. The method is applied to flatness pattern recognition. Denote seven kinds of basis flatness modes that meet with usually by Legendre orthodoxy polynomials, don＇ t need more sample in training the network. It is a more simple and availability method in flatness pattern recognition and makes reference for carrying out the flatness control. Simulation result shows this way is practicable.

关 键 词：离散Hopfield神经网络(DHNN) 随机扰动优化设计 勒让德多项式 板形模式 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]