机器学习中谱聚类方法的研究

Research on Spectral Clustering in Machine Learning

最近几年,谱聚类方法在模式识别中得到了广泛的应用。与传统的聚类方法比较,它具有能在任意形状的样本空间上聚类,且收敛于全局最优解的优点。本文着重介绍了谱方法的基本原理、相应的算法、研究状况及其在模式识别领域中的应用,同时指出了它的关键问题与未来的研究方向。

Recently spectral clustering has wide application in pattern recognition. Comparing with traditional clustering methods, it has ＂global＂ optima solution. This paper introduces the principle, algorithms of spectral clustering, and the application in pattern recognition. And also it points out the key problems and future direction.