摘要
在局部线性嵌入算法(LLE)中寻找最优近邻数常用试凑法进行搜索,需要大量的时间才能得到最优结果.为此提出基于自适应近邻的局部线性嵌入算法(ANLLE),算法首先给出一个相似性度量函数,然后据此为各个样本设定阈值,根据每个样本周围数据分布情况,为每个样本自动设置不同近邻数,最后在各个样本近邻数不相同情况下进行数据降维及待测样本的分类.在人脸数据库及手写数字数据库上的对比实验表明,ANLLE算法识别性能高于标准LLE算法及邻域线性嵌入算法(NLE).
Finding the optimal neighbors in a locally linear embedding（LLE） algorithm is still an unresolved problem.Trial and error,a commonly used method,requires much time to obtain the optimal result.In this paper,an adaptive neighborhoods based locally linear embedding algorithm（ANLLE） was proposed which first provided a new similarity measure function and secondly set a threshold for each sample.Then,it set various numbers of neighbors for each sample according to different distributions around it.Finally,the ANLLE reduced the dimension of samples and classified them to be tested in the case of different neighbors for each sample.A comparison of ANLLE,the neighborhood linear embedding algorithm（NLE）,and a standard LLE algorithm in human faces and script databases proves that the ANLLE is more effective than standard LLE and NLE algorithms.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2012年第4期489-495,共7页
Journal of Harbin Engineering University
基金
国家自然科学基金资助项目(60873038)
关键词
局部线性嵌入
自适应近邻
维数约减
嵌入算法
最优近邻
相似性度量函数
locally linear embedding
adaptive neighborhoods
dimensionality reduction
embedding algorithm
optimal neighbors
similarity measure function
