Discriminative Semi-Supervised Clustering Analysis with Pairwise Constraints

Most existing semi-supervised clustering algorithms with pairwise constraints neither solve the problem of violation of pairwise constraints effectively, nor handle the high-dimensional data simultaneously. This paper presents a discriminative semi-supervised clustering analysis algorithm with pairwise constraints, called DSCA, which effectively utilizes supervised information to integrate dimensionality reduction and clustering. The proposed algorithm projects the data onto a low-dimensional manifold, where pairwise constraints based K-means algorithm is simultaneously used to cluster the data. Meanwhile, pairwise constraints based K-means algorithm presented in this paper reduces the computational complexity of constraints based semi-supervised algorithm and resolve the problem of violating pairwise constraints in the existing semi-supervised clustering algorithms. Experimental results on real-world datasets demonstrate that the proposed algorithm can effectively deal with high-dimensional data and provide an appealing clustering performance compared with the state-of-the-art semi-supervised algorithm.