A feasible solution model for manifold clustering

For manifold clustering, it is very time-consuming to obtain the optimal partition. To reduce time complexity, this paper introduces a feasible solution method, which is able to guarantee optimal solution and improve the efficiency of the k-way partition model. Through reasoning the constraint conditions that clustering feasible solutions subject to the minimum total cost, a new sub-cluster with high cohesion is suggested. Further, a clustering framework based on sub-clusters is proposed, and a subcluster-based optimal partition model (SB-OPM) is given. For complex manifold data, the proposed model can not only ensure the global optimal partition (minimum total cost) but also reduce computational costs. In particular, this model has shown excellent performance in comparative experiments and Friedman test.