基于文化粒子群算法的KPCA特征提取  

KPCA feature extraction based on CBPSO algorithm

作　　者：赵敏[1] 杨恢先[2] 欧训勇[3] 

机构地区：[1]湘潭大学信息工程学院,湖南湘潭411105 [2]湘潭大学材料与光电物理学院,湖南湘潭411105 [3]琼州学院物理系,海南五指山572200

出　　处：《计算机应用研究》2009年第8期2908-2911,共4页Application Research of Computers

基　　金：海南省自然科学基金资助项目(60897)

摘　　要：如何选择最优或接近最优的核函数使分类错误率降低,是KPCA应用于特征提取的关键。为了优化核函数,提高特征提取的能力并降低分类错误率,在研究了文化算法(cultural algorithm,CA)、粒子群优化(particleswarm optim ization,PSO)相关文献的基础上,提出了一种文化粒子群算法(cultural based PSO,CBPSO)流程,并将此算法用于训练核函数参数,实现了KPCA和CBPSO的集成,有效地提高了核函数的优化选择。通过比较CBPSO-KPCA与GA-KPCA的仿真结果,表明该方法具有较好的结果和更少的计算量,克服了GA(genetic algo-rithm)的缺点。How to choose the best or near kernel function to reduce test error rate is the key of KPCA applied to feature extraction. In order to the optimization of kernel function, increased the ability of feature extraction and decrease the test error rate, on the basis of research of CA, PSO, this paper proposed a program flow of CBPSO used for training kernel function and built CBPSO-KPCA. This approach could effectively optimize kernel function. Compared CBPSO-KPCA simulation results with GAKPCA simulation results, it shows that CBPSO produces highly competitive results at a relatively low computational cost, overcomes the disadvantage of GA.

关 键 词：文化算法 粒子群优化 文化粒子群算法 核主分量分析 特征提取 遗传算法 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]