支持向量机原始问题研究综述  

Research of support vector machine in the primal

作　　者：赵春婕[1] 王树勋[1] 

机构地区：[1]陕西理工学院数学系,陕西汉中723001

出　　处：《陕西理工学院学报（自然科学版）》2010年第2期58-64,共7页Journal of Shananxi University of Technology:Natural Science Edition

基　　金：陕西省教育厅自然科学基金项目(07JK209)

摘　　要：支持向量机学习器往往是通过求解原二次优化问题的对偶问题获得的。诸多研究表明,支持向量机原始问题同样可以适当地处理约束项,同时,突破以前原二次优化问题不能利用核函数的认识误区,通过引入核函数建立一个无约束优化问题,利用传统优化方法进行求解。理论分析和实验表明,支持向量机原始问题也能实现对数据的高效学习,而且在大规模数据学习问题上,较之求解对应的对偶问题获得的近似解更可靠,参数选择也更好进行。Many algorithms about support vector machine gained the optimal solution through solving corresponding dual programming over the last decade.Recently,many papers about support vector machine in primal space were published,likewise,constraint items were solved efficiently and a unconstrained model was established by introducing the kernel function in primal problem which could be solved by using traditional optimization methods.Based on theoretical analysis and experimental results,it was effective,especially the approximation solution was superior to that to the dual problems when encountering large scale problem.Moreover,the problem of parameter selection could be completed easily.

关 键 词：向量机 学习器 优化 核函数 

分 类 号：TP391.6[自动化与计算机技术—计算机应用技术]