K-means clustering with manifold

K-means clustering is a popular conventional clustering algorithm. As it does not use the structure information of data sets, sometime the clustering result will be dissatisfied. Manifold learning algorithms can reveal the low-dimensional geometry structure of the data sets. In this paper, we combine K-means clustering algorithm with manifold learning algorithms into a coherent framework. We show the proposed algorithms KCM(K-means clustering with manifold) approaches can obtain good clustering results on UCI data sets. We also illustrate that the KCM clustering algorithms can be naturally extended to semi-supervised clustering. Experimental results also show the effectiveness of the semi-supervised clustering approaches.