摘要
传统流形学习算法虽然是一种常用的有效降维方法,但由于其自身计算结构的限制,往往存在数据分析不足和计算时间较长等问题.为此提出一种基于谱聚类的流形学习算法(spectralclustering locally linear embedding,SCLLE),并对其机理以及优点给予了实例证明.在UCI和NCBI数据集上的实验结果表明,该算法具有较好的识别效果和计算性能.
Although traditional manifold learning algorithms are common and effective dimension reduction methods, they still have calculating structure limits of their own, which lead to some problems such as inadequate data analyses and long calculation time. Therefore, on the basis of spectral clustering, a manifold learning algorithm named SCLLE （spectral clustering locally linear embedding） was proposed and its mechanisms as well as its advantages were demonstrated. Experiments with UCI and NCBI data sets show that the proposed algorithm has better recognition effect and computational performance.
基金
国家自然科学基金重点项目(71131002)
国家自然科学基金面上项目(71071045
71001032
70801024)
中国博士后科学基金项目(20110490831)资助
关键词
谱聚类
流形学习
SCLLE算法
spectral clustering
manifold learning
spectral clustering locally linear embedding algorithm
