摘要
支持向量域数据描述(SVDD)是一种单值分类算法,用于将目标样本与其他非目标样本区分开来。本文引入数学中曲率的概念,根据分类边界线附近支持向量曲率的大小来对训练集进行约减;提出了一种约减型的支持向量域数据描述快速训练算法FSVDD,该算法与传统SVDD相比减少了训练时所需的支持向量数目,因而训练时间极大减少,同时分类性能几乎不受大的影响,该算法在大规模训练样本学习中具有现实意义.
As a type of one-class classification algorithm,Support Vector Data Description(SVDD) was used to distinguish target objects from outlier objects.In order to accelerate its classification speed when it faced with large scale classification problems,It introduced the mathematics concept of curvature,and reduced the training samples according to the curvature value of support vectors locating on the classification boundary.then a fast learning SVDD(FSVDD) algorithm based on the reduced support vectors set was presented.Compared with the traditional SVDD,FSVDD only uses reduced support machines to construct the final classification boundary,so the training time is decreased greatly,meanwhile,the classification performance of the FSVDD has no obvious loss.The experimental results show that the proposed algorithm is practical and effective.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2006年第z1期798-800,共3页
Chinese Journal of Scientific Instrument
关键词
支持向量域数据描述
支持向量机
快速学习
support vector data description support vector machine fast learning
