局部线性重构分类算法研究  

ON CLASSIFICATION ALGORITHM BASED ON LOCALLY LINEAR RECONSTRUCTION

作　　者：李长彬[1] 黄东[1] 黎永壹[1] 

机构地区：[1]钦州学院数学与计算机科学学院,广西钦州535000

出　　处：《计算机应用与软件》2013年第4期156-158,210,共4页Computer Applications and Software

基　　金：广西自然科学基金项目(09236003)

摘　　要：局部线性重构(LLR)是局部线性嵌入算法(LLE)的一个基本步骤,其目的是用线性重构的方法来表达向量数据之间的局部邻域关系。提出一种快速的监督分类算法,它采用训练集中的向量数据重构测试数据,能够最好地重构给定测试数据的类别被判定为待求标签。与相关算法的实验比较表明,该算法在分类准确性和计算时间上均有明显的优势。Locally linear reconstruction（LLR） is a fundamental step in dimensionality reduction algorithm known as locally linear embedding（LLE） for representing the local neighbourhood relationship between vector data through linear reconstructions.We present a fast supervised classification algorithm,it reconstruct the test data with vector data in training set and can best reconstruct the judgement of the class of given test data as the estimating label.Comparison with the experiment of correlated algorithm shows that the proposed algorithm clearly outperforms in classification accuracy and computing time.

关 键 词：监督分类 局部线性重构 重构误差 降维 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]