摘要
采用混合多流形谱聚类模型(SMMC)对独立子空间、非独立子空间,非线性良分离及非线性交叉等流形聚类中的四种典型数据进行聚类,并与其他流形聚类方法进行比较,发现SMMC模型聚类效果良好且具有强鲁棒性和泛化能力.将SMMC模型运用于具有混合多流形结构的工件外部边缘轮廓进行聚类,结果显示SMMC模型能够很好的将其分为三类.针对SMMC模型复杂度高、选取参数困难及运行时间长的问题,提出了基于模拟退火遗传算法SMMC模型,结果发现改进后的模型能够大大缩短运行时间.
This paper firstly adopts spectral multi-manifold clustering （SMMC） to four typical data of independent subspace, non-independent subspace, nonlinear well-separated and nonlinear overlapped, then follows by a comparison in results among these four clustering methods. The research shows that the SMMC works fine. Moreover, there are strong robustness and well generalization in it. Afterwards, the SMMC is used to cluster in contours of a workpiece characterized by multi-manifold, which is well classified into three groups. Aimed at the problems of high complexity in time and space as well as difficulty in parameters selection, we put forward a modified SMMC model with the advantages of shortening the uptime greatly based on simulated annealing genetic algorithm.
出处
《数学的实践与认识》
北大核心
2016年第14期163-172,共10页
Mathematics in Practice and Theory
关键词
大数据
流形学习
多流形谱聚类
模拟退火遗传算法
big data
manifold learning
spectral multi-manifold clustering
simulated annealing genetic algorithm
