Experimental Comparison of Support Vector Machine Training Algorithms
支持向量机训练算法的实验比较

Support vector learning algorithm is based on structural risk minimization principle.It combines two remarkable ideas: maximum margin classifiers and implicit feature spaces defined by kernel function.Presents a comprehensive comparison of three mainstream learning algorithms: SVM~(light),Bsvm,and SvmFu using face detection,MNIST,and USPS hand-written digit recognition applications.