基于万有引力定律的分类方法研究  

AN EXPLORATION OF CLASSIFICATION BASED ON THE LAW OF UNIVERSAL GRAVITATION

作　　者：杜隆胤[1] 

机构地区：[1]贵阳学院计算机科学系,贵州贵阳550003

出　　处：《计算机应用与软件》2013年第2期205-207,共3页Computer Applications and Software

基　　金：贵州省教育厅自然科学基金项目(20090060);贵州省科学技术基金项目(20102129)

摘　　要：在模式识别研究中,大多利用发展成熟的数学理论或工具进行分类器设计。运用联系的观点,将类聚问题类比于宇宙中各星系的形成,把物理知识迁移到分类运算中,设计一种参照物理过程的分类方法——基于万有引力定律的分类方法。即通过对比特征向量与各训练集中类间的"引力"来决定其分类。通过大量实验证明该分类方法运算量小、具有很好的鲁棒性。In pattern recognition researches,the designs of classifiers mostly employ the mature-developed mathematical theories or tools.In this paper we look for other way,use the viewpoint of connection to compare the clustering problem to the formation of the galaxies in universe,and by translating physical knowledge to classification operations,we design a kind of classification method with reference to physical process – the classification method based on the Law of Universal Gravitation.That is,it determines the classification by comparing eigenvectors and interclass ＂gravitation＂ between each training class.Numerous experiments have proved that the computation of the classification method is moderate with good robustness.

关 键 词：模式识别 万有引力 特征向量 分类器 鲁棒性 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]