摘要
主要针对几种典型数据的多流形结构分析问题进行了研究.综合分析多种谱聚类算法优缺点,以谱多流形聚类算法为主线,结合实验结果对多种谱聚类算法进行了分析,最后针对数据空间密度不均匀的情况对谱多流形聚类算法进行了一定的改进,提出了一种基于自适应近邻值的谱多流形聚类算法,并通过实验证明其达到了混合多流形聚类的目的.
This article mainly aims at the analysis of the manifold structure with several typical data. Comprehensive analysis on advantages and disadvantages of various spectral clustering algorithm had been clone which based on manifold spectral clustering algorithm as the main line. Combining with the experimental results, a variety of spectral clustering algorithm is analyzed, Finally, in the uneven density data space, an improved algorithm was put forward a kind of based on adaptive neighbor value spectrum manifold clustering algorithm, and it was proved to be useful through the experiment.
出处
《数学的实践与认识》
北大核心
2016年第14期173-179,共7页
Mathematics in Practice and Theory
关键词
谱聚类
流形学习
自适应近邻值
spectral clustering
manifold learning
adaptive neighbor values