统计学习理论及支持向量机概述

Introduction Statistical Learning Theory and Support Vector Machines

针对传统统计模式识别理论中基于大数定理的假设,介绍了统计学习理论和以该理论为基础的支持向量机模式识别方法。指出了以结构风险最小化为原则的分类器设计方法,即同时兼顾分类能力最优化和经验风险最小化。支持向量机是统计学习理论的VC维理论和结构风险最小原理的具体实现,他通过非线性变换将输入空间变换到一个高维空间,然后在这个新空间中求取最优线性分类面。

The statistic learning theory and the method of supporting vector mechanism pattern identification base on it are described aiming to the hypothesis of big number theorem in traditional pattern identification theory. Presents the design methods of classification implement according to the minimum structure risk, namely according to both optimization of classification ability and minimum of experience risk. The supporting vector mechanism is the concrete realization of VC dimension theory of statistic learning theory and minimum structure risk principle which transform the income space into a high dimension space through non -linearity transformation and seek the optimum linearity classification facet.