摘要
为了加快并行下降方法(CD)用于线性支持向量机(SVM)时的最终收敛速度,将Rosenbrock算法(R)用于线性SVM.在内循环,R通过解一个单变量子问题来更新w的一个分量,并同时固定其他分量不变;在外循环,采用Gram-schmidt过程构建新的搜索方向.实验结果表明,与CD相比,R加快了最终的收敛,在分类中能更快地获得更高的测试精度.
To improve the speed of final convergence of coordinate descent method (CD) when it is applied to linear support vector machine (SVM), Rosenbrock algorithm (R) is applied to linear SVM. In inter iterations, R updates one component of w by solving a one-variable sub-problem while fixing other components. In outer iterations, the new search direction is constructed by the Gram-schmidt procedure. Experimental results show that, compared with CD, R aceerlates final convergence and achieves higher testing accuracy more auicklv in classification
出处
《控制与决策》
EI
CSCD
北大核心
2009年第12期1895-1898,共4页
Control and Decision
基金
国家自然科学基金项目(60574075
60705004)