Soft-balanced CAN clustering method

As one of the most outstanding clustering methods, CAN can obtain accurate clustering results by imposing rank constraint to the Laplacian matrix. However, CAN ignores the balanced distribution of real-world sample data. To overcome the disadvantage, we propose a clustering method, referred to as Soft-balanced CAN clustering method (SBCAN) in this paper by introducing a soft-balanced clustering regularizer. Since our proposed method can perform two steps simultaneously with soft-balanced constraint: similarity measurement and data clustering, it can obtain the optimal clustering result. The experiments on six real-world benchmarks demonstrate that our proposed method has superior performance than the state of the art.