RBF核SVM及其应用研究  

SVM with RBF kernel and its application research

作　　者：燕孝飞[1,2] 葛洪伟[1,2] 颜七笙[1,2] 

机构地区：[1]江南大学信息工程学院,江苏无锡214122 [2]东华理工学院,江西抚州344000

出　　处：《计算机工程与设计》2006年第11期1996-1997,2011,共3页Computer Engineering and Design

摘　　要：因其核函数的良好性态,RBF核SVM(RBF-SVM)在实际应用中表现出良好的学习性能,但是RBF核函数中的参数对SVM的性能起决定性作用。阐述了RBF-SVM的性能随着变化而变化的规律,并将RBF-SVM引入自动羽绒识别系统中。根据自动羽绒识别系统的实际需求和RBF-SVM的性能变化规律,论述了本系统中参数的选取依据和选取过程,并且给出了的相关曲线变化图。通过研究,最后得到适合本系统的识别模型,从而提高了系统的总体识别率。同时,也验证了RBF-SVM的良好特性和其受参数的约束规律。Because of good properties of RBF kernel, SVM with RBF kernel （RBF-SVM） shows good learning performance in the practical application. But the performance of RBF-SVM is influenced greatly by the scale parameter. How the affects the function of RBF-SVM is expatiated. Then RBF-SVM is introduced to feather and down category recognition system. Based on the practical of the system and the changing performance of the RBF-SVM, the principle and the process of the parameter are addressed. The picture of the changing curve of is given. A RBF-SVM model of the recognition system is presented via study, which increases the recognition rate of the system. Simultaneously the good learning performance of RBF-SVM and its restriction law are validated.

关 键 词：支持向量机 径向基核函数 学习性能 羽绒识别 变化曲线 识别率 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]