对角矩阵指数优化的局部保持映射算法  

Improved local preserving projection algorithm based on exponential diagonal matrix

作　　者：安亚静[1] 王士同[2] 

机构地区：[1]江南大学物联网工程学院,江苏无锡214122 [2]江南大学数字媒体学院,江苏无锡214122

出　　处：《计算机工程与应用》2011年第36期197-202,225,共7页Computer Engineering and Applications

基　　金：国家自然科学基金(the National Natural Science Foundation of China under Grant No.60704047);国家自然科学基金重大研究计划(No.9082002)

摘　　要：局部保持映射(LPP)算法利用欧几里德距离求得权值累加得到对角矩阵,利用结果进行降维。对于这个算法是否可以进一步优化还值得进一步探讨。对该算法所依据的公式进行修改,在对角矩阵上引入指数参数,形成对角距阵指数优化的局部保持映射算法。通过实验可以证明,对角距阵指数优化的局部保持映射算法能够影响降维的结果,可以使得降维更容易得到接近本征维数的投影向量,通过实验验证降维后的识别效果和对噪声的敏感度。Local Preserving Projection algorithm（LPP） obtains the diagonal matrix by computing the sum of the Euclidean distance and it can make use of the results to reduce dimensions.However,it is worth to do further investigation whether the algorithm can be optimized.In the paper,some modification are made to the formula which the algorithm rests on.With adding the exponential parameter,the improved local preserving projection algorithm based on exponential diagonal matrix can be got.Through a lot of experiments it has an affect on the results of the dimensionality reduction.It is easier to approach the intrinsic dimension of the datas.It also checks the discrimination after reducing the dimensions and the sensitivity of noise.

关 键 词：维数约简 局部保持映射(LPP) 对角矩阵 指数参数 噪声 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]