线性支持向量分类机优化问题解的二阶充分条件  

Second order sufficient condition for optimization problem of a linear support vector classifier

作　　者：蔡春[1] 刘宝光[2] 邓乃扬[3] 

机构地区：[1]北京联合大学应用文理学院,北京100083 [2]北京理工大学理学院,北京100086 [3]中国农业大学理学院,北京100083

出　　处：《中国农业大学学报》2006年第6期92-95,共4页Journal of China Agricultural University

基　　金：国家自然科学基金资助项目(10371131)

摘　　要：优化问题的二阶充分条件是研究灵敏度分析的基础,支持向量机是数据挖掘的新方法。针对线性支持向量分类机优化问题,研究了其解的二阶充分条件,给出了二阶充分条件成立的假设条件。研究表明,该假设条件很弱,用支持向量机算法求解实际问题时,通常假定这一条件成立,特别地,对线性可分支持向量分类机优化问题,其解一定满足这一条件,满足二阶充分条件成为当然成立的事实。A second-order-sufficient-condition is one of basis for sensitivity analysis of optimization problems. A hypothesis to ensure the tenability of a second-order-sufficient-condition of a linear support vector classifier is presented in the paper, which is a weak one for the solution of a linearly separable support vector classifier must meet it. In addition, the problem of a linear support vector classifier is solved under such a hypothesis.

关 键 词：支持向量分类机 二阶充分条件 数据挖掘 起作用约束 

分 类 号：TP13[自动化与计算机技术—控制理论与控制工程]