支持向量预选取的K边界近邻法
Pre-extracting support vectors for Support Vector Machine using K nearest bound neighbor method
李庆 1胡捍英1
作者信息
- 1. 解放军信息工程大学通信工程系,河南郑州450002
- 折叠
摘要
支持向量机是基于统计学习理论的一种新兴的模式识别方法,在解决小样本、非线性及高维模式识别问题中表现出了突出的优势.但其支持向量的选取相当困难,这也成为限制其应用的瓶颈问题.本文提出了一种支持向量预选取的方法-K边界近邻法.该方法能有效提取包含支持向量的边界向量机,在不影响分类性能的情况下,极大减少了训练样本,提高训练速度.且新方法避免了数据分布的影响及对先验知识的依赖.仿真实验证明了该方法的可行性和有效性.
Abstract
Support Vector Machines (SVMs) is a novel pattern recognition method based on statistic learning theory,which shows prominent advantages in solving small samples,non-linear and high dimension problems.However,the selection of support vectors (SVs) is quite difficult and time-consuming,which becomes a bottleneck of the application.A new method called K nearest bound neighbor is proposed for pre-extracting support vectors.It greatly reduces the training samples and speeds up the SVM training without any loss of classification performance.Also the method avoids the impacts of the sample distribution and the dependence on prior knowledge.Our experiments shows remarkable results to support our idea.
关键词
支持向量机/K边界近邻法/预选取/边界向量/支持向量Key words
Support Vector Machines/K nearest bound neighbor method/pre-extraction/bound vectors/Support Vectors引用本文复制引用
基金项目
国家重大专项资助项目(2009ZX03003-007)
出版年
2013
NETL
NSTL
维普
万方数据