SVM(support vector machines)是一种基于结构风险最小化原理的分类技术.给出实现结构风险最小化原理(最大边缘)的另一种方法.对线性可分情形,提出一种精确意义下的最大边缘算法,并通过闭凸包收缩的概念,将线性不可分的情形转化为线性可分情形.该算法与SVM算法及其Cortes软边缘算法异曲同工,但理论体系简单、严谨,其中的优化问题几何意义清楚、明确. SVM (support vector machines) is a classification technique based on the principle of structural risk minimization, and presents another method to realize the principle of structural risk minimization (maximum margin) .For linear separable cases, Maximum edge algorithm, and the concept of closed convex hull contraction, the linear inseparable case is transformed into a linear separable case.The algorithm is similar to the SVM algorithm and its Cortes soft edge algorithm, but the theoretical system is simple and rigorous, in which the optimization problem geometry Meaning is clear and clear.