ACTIVE CONSTRAINTS SELECTION BASED ON DENSITY PEAK ESTIMATION

Abstract

Semi-supervised clustering, which integrates side information from users to enhance clustering performance, has gained considerable attention in the research community. However, the quality of clustering is highly dependent on the side
information provided, and different inputs can lead to different results. In this paper, we propose an active learning approach for selection good constraints, which employs a min-max strategy and density-based estimation of data points to optimize the constraints selection process. Experimental evaluations on datasets from UCI and an real face image data show the effectiveness of our method.