基于核向量机的模式分类及其实验测试
Pattern Classification and Experiment Testing Based on Core Vector Machine
-
摘要: 文中使用一种新的SVM变种——核向量机来对大样本数据集进行训练建模,进而求解模式分类问题.CVM算法是将核函数转换为最小包围球问题进行求解,可以解决任何线性或非线性分类问题.测试结果表明,核向量机可以快速对大样本数据进行分类并能产生较少的支持向量.Abstract: To handle such kind of large datasets,we use a new kind of CVM variable--Core Vector Machine(CVM)for pattern classification.CVM algorithm formulates original kernel methods in SVM as a Minimum Enclosing Ball(MEB) problems in computational geometry and it can be used with any linear/nonlinear kernels.Experiment shows that the CVM is as accurate as existing SVM implementations,but it gets smaller support vectors and much faster than SVM for very larger datasets.