JOINT CLUSTERING OF SPARSE NEIGHBORHOOD CONSTRAINTS BASED ON CLASS CONSISTENCY LEARNING

In order to fully mine the feature structure and improve the clustering performance, a joint clustering method with sparse neighborhood constraints based on category consistency learning is proposed. The joint clustering problem was transformed into a tri-factorization of nonnegative matrix with dual regularizer. Based on the nonnegative matrix decomposition, two regularizers were added to make the data relevance consistent with the label assignment. A multiplication alternation scheme for objective optimization was introduced, and the convergence and correctness of the algorithm were proved theoretically. The three evaluation methods were verified on six data sets, and their parameter sensitivity was analyzed. Experiments show that the proposed algorithm has better performance.