计算机技术与发展2017，Vol.27Issue(1)：61-64,69,5.DOI:10.3969/j.issn.1673-629X.2017.01.014
基于局部信息融合的正交稀疏保留投影分析
Analysis of Orthogonal Sparse Preserving Projection Based on Local Information Fusion
摘要
Abstract
There are many sample classification criterion in the field of pattern recognition,and the mostly used one recently is to maintain the sparse reconstruction relationship of original data samples to the sample space after projection transformation so as to increase the ac-curacy of classification. SPP is a typical algorithm based on the idea. In finding the optimal projection transformation,SPP is based on the global point view,however the spatial structure of the sample is nonlinear and the local linear structure is not considered. At the same time,the SPP algorithm is not orthogonal,and there is redundant information between the feature transform. Based on the above shortcom-ings,orthogonal sparse preserving projection based on local information fusion is proposed,and the orthogonality and the local structure information of samples are merged into the SPP algorithm. At the same time,the proposed algorithm is validated on the AR and CAS-PEAL face database.关键词
稀疏保留投影/局部近邻信息/正交性/迭代终止准则Key words
sparsity preserving projections/local neighbor information/orthogonality/iterative stopping criterion分类
信息技术与安全科学引用本文复制引用
袁安鼎,荆晓远,吴飞..基于局部信息融合的正交稀疏保留投影分析[J].计算机技术与发展,2017,27(1):61-64,69,5.基金项目
国家自然科学基金资助项目(61272273) （61272273）
江苏省普通高校研究生科研创新计划(CXLX13465) （CXLX13465）
