Component?based nearest neighbour subspace clustering

In this paper, the problem of clustering data points that lie near or on a union independent low-dimensional subspaces is addressed. To end, popular spectral clustering-based algorithms usually follow two-stage strategy initially builds an affinity matrix and then applies clustering. However, inappropriate does not sufficiently connect lying same subspace will easily lead to issue over-segmentation. alleviate issue, building based hypotheses generated by iterative sampling operation according Random Cluster Model under framework energy minimisation proposed. Specifically, each hypothesis from large number component in K-nearest neighbour graph. Extensive experiments synthetic real-world datasets show proposed method can improve connectivity provide competitive results against state-of-the-art methods.