η-one-class问题和η-outlier及其LP学习算法  

η-One-Class Problems and η-Outliers with Their LP Learning Algorithms

作　　者：陶卿[1] 齐红威[2] 吴高巍[2] 章显[1] 

机构地区：[1]中国人民解放军炮兵学院,合肥230031 [2]中国科学院自动化研究所,北京100080

出　　处：《计算机学报》2004年第8期1102-1108,共7页Chinese Journal of Computers

基　　金：国家自然科学基金项目 ( 60 175 0 2 3 );安徽省自然科学基金项目( 0 10 42 3 0 4);安徽省优秀青年科技基金项目资助

摘　　要：用SVM方法研究one class和outlier问题 .在将one class问题理解为一种函数估计问题的基础上 ,作者首次定义了 η one class和 η outlier问题的泛化错误 ,进而定义了线性可分性和边缘 ,得到了求解one class问题的最大边缘、软边缘和ν 软边缘算法 .这些学习算法具有统计学习理论依据并可归结为求解线性规划问题 .算法的实现采用与boosting类似的思路 .实验结果表明该文的算法是有实际意义的 .One-class and outlier problems are investigated by using the idea of Support Vector Machines. Based on regarding a one-class problem as the one to estimate a function, the generalization error for the one-class problem is defined for the first time. The linear separability, margin and optimal linear classifier are then defined and the regular SVM is reformulated into a framework for one-class problems. Each of the linear algorithms is motivated theoretically and they can be formulated as some linear programming problems. The proposed algorithms can be implemented by the techniques in boosting algorithms. Some synthetic and real experiments illustrate that the algorithms in this paper are practical and effective.

关 键 词：one-class问题 OUTLIER 最大边缘 统计学习理论 支持向量机 线性规划问题 BOOSTING 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]