基于局部重构与全局保持的半监督维数约减算法  

Local Reconstruction and Global Preserving Based Semi-supervised Dimensionality Reduction Algorithm

作　　者：韦佳[1] 文贵华[1] 王文丰[2] 王家兵[1] 

机构地区：[1]华南理工大学计算机科学与工程学院,广州510006 [2]南昌工程学院信息工程学院,南昌330099

出　　处：《计算机科学》2011年第8期201-204,共4页Computer Science

基　　金：国家自然科学基金(60973083);华南理工大学中央高校基本科研业务费专项资金(2009ZM0189;2009ZM0175)资助

摘　　要：针对基于局部与全局保持的半监督维数约减算法(LGSSDR)对邻域参数选择比较敏感以及对邻域图边权值设定不够准确的问题,提出一种基于局部重构与全局保持的半监督维数约减算法(LRGPSSDR)。该算法通过最小化局部重构误差来确定邻域图的边权值,在保持数据集局部结构的同时能够保持其全局结构。在Extended YaleB和CMU PIE标准人脸库上的实验结果表明LRGPSSDR算法的分类性能要优于其它半监督维数约减算法。Considering that Local and Global Preserving Based Semi-Supervised Dimensionality Reduction（LGSSDR） is sensitive to the selection of neighborhood parameter and inaccurate in the setting of the edge weights of neighborhood graph,a new algorithm of Local Reconstruction and Global Preserving Based Semi-Supervised Dimensionality Reduction（LRGPSSDR） was proposed in this paper.The algorithm can set the edge weights of neighborhood graph through minimizing the local reconstruction error and can preserve the global geometric structure of the sampled data set as well as preserving its local geometric structure.The experimental results on Extended YaleB and CMU PIE face database demon-strate that LRGPSSDR is better than other semi-supervised dimensionality reduction algorithms in the performance of classification.

关 键 词：边信息 局部重构 半监督学习 维数约减 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]