On Kernelized Sequential Hard Clustering
Yukihiro Hamasuna, Yasunori Endo · 2016
A method of sequential clustering extracts a cluster sequentially without determining the number of clusters. The sequential hard clustering is based on noise clustering and one of the typical sequential clustering methods. A kernelized sequential hard clustering is proposed by introducing the kernel method to sequential hard clustering to handle datasets which consists non-linear clusters and execute robust clustering. The performance of the proposed method is evaluated with a typical dataset which consists non-linear cluster boundary. Negative results are obtained through numerical examples and those show that the proposed method can not extract non-linear clusters sequentially.