神经网络在环境质量评价中的问题  

Problem of Neural Network in Assessment of Environmental Quality

作　　者：冯利华[1] 张行才[1] 

机构地区：[1]浙江师范大学地理系,金华321004

出　　处：《微计算机应用》2008年第2期46-49,共4页Microcomputer Applications

基　　金：国家自然科学基金资助项目(40771044);浙江省科技计划重点项目(2006C23066);义乌市科技攻关项目(2006-G3-20)

摘　　要：在环境质量评价中,神经网络被认为是一种客观而实用的评价方法。因为神经网络的最大特点在于不需要设计任何数学模型,具有自学习、自组织、自适应和容错性等优点,因而在模式识别和系统辨识等领域获得了许多成功的应用。但数据分析表明,网络输出结果并不完全符合实际情况,有时甚至严重失真,存在着一系列问题,而这些问题是由神经网络本身的性质所决定的。了解了这些问题,就能够使我们对神经网络计算的环境质量等级分类有更全面和深刻的认识。Neural network was taken for an objective and practical evaluation method in assessment of environment quality. The advantage of neural network is that there is no need to devise a mathematical model. It has the advantages of self - learning, self - organization,self- adaptation and fault tolerance. Thus it is successfully applied to pattern recognition and system identification etc. But datum analysis demonstrated that all the results exported by neural network didn＇t conform actual conditions. Sometimes it wasn＇t true, and there were a lot of problems. These problems depended on the property of neural network itself. Studying of these problems increases our understanding of grade classification of environmental quality calculated by neural network.

关 键 词：神经网络 等级分类 环境质量评价 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] X820[自动化与计算机技术—控制科学与工程]