机构地区: 华南师范大学计算机学院
出 处: 《广州大学学报(自然科学版)》 2007年第4期44-48,共5页
摘 要: 在模式识别领域,基于Fisher判别准则的Foley-Sammon变换技术有很大的影响.但是线性判别并不总是最优的.文章提出了一种基于核技巧(Kernel tricks)的非线性的特征提取技术KFST(Foley-Sammon Transformwith Kernels)——通过引入核技巧,可以在特征空间中有效计算FST.特征空间中的线性特征提取对应于输入空间的非线性特征提取.试验表明,KFST比FST具有更好的特征提取能力. Fisher discriminant based Foley-Sammon Transform has great influence in the area of pattern recognition. Linear discriminants are not always optimal, so a new nonlinear feature extraction method based on kernel trick is presented in this paper. The kernel trick amounts to performing the same algorithm (FST) in feature space. Linear feature extraction in feature space corresponds to nonlinear feature extraction in input space. Experimental results show that, compared with FST, KFST has more powerful ability of feature extraction.
领 域: [自动化与计算机技术] [自动化与计算机技术]