基于多视图树的多关系分类  

Multi-relational Classification Based on Multiple View Tree

作　　者：郑利雄[1] 陈琼[1] 沈勇明[1] 

机构地区：[1]华南理工大学计算机科学与工程学院,广州510006

出　　处：《计算机工程》2010年第19期96-98,共3页Computer Engineering

基　　金：国家自然科学基金与中国民用航空总局联合基金资助项目(60776816);广东省自然科学基金资助重点项目(8251064101000005)

摘　　要：提出一种新型的利用多视图树的方法解决多关系分类的问题。在多视图分类方法中,不同视图对分类任务贡献不同,视图间存在互补性。对视图互补性进行研究,提出视图互补性的度量方法。根据视图互补性选择视图进行集成,构造多视图树,用于多关系分类。实验结果表明,多视图树无论在分类效果还是效率上,都优于现有的多关系分类方法。In this paper,a new approach using multiple view tree is proposed to solve the problem of multi-relational classification.In the multiple view classification,different individual view has different contribution to the classification task.Views are complementary to the classification.The complementary property of views is studies and a compute method of views is proposed complementary.The views are chosen to integrate to construct a multi-view tree according to the views complementary.Experimental results show that multi-relational classification method based on multiple view tree outperforms the existing methods in terms of accuracy and efficiency.

关 键 词：多关系分类 多视图 二叉树 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]