The K-Exemplars Clustering Method
Hsiou Hen Kao, Li Huang, Miin Jye Wen, Kuo Lung Wu · Applied Mechanics and Materials · 2013
In order to apply the concepts of k-means to deal with any specified dissimilarity measures, we propose a k-exemplars clustering method that modifies k-means by restricting the cluster centers on data points. The proposed method not only has similar clustering accuracy as k-means but also faster convergence. According to the definition of the objective function of k-exemplars, the proposed method can be used to deal with a relational data set, and the cluster centers (exemplars) of each cluster will also be extracted. Hence, the k-exemplars can work in an environment with specified dissimilarity measures.