MVS-based semi-supervised clustering

Abstract

Semi-supervised clustering is a popular machine learning technique, used for challenge data categorization tasks, when some prior knowledge is available to users. In this paper, we report the empirical studies on our newly proposed semi-supervised clustering framework, which utilizes multiple viewpoints for the similarity measure, with the help of the prior knowledge. Two different MVS-based approaches are developed for knowledge given in either class labels or pair-wise constraints, namely LMVS and PMVS respectively. Extensive experimental studies performed on a few benchmark datasets demonstrate the effectiveness of the proposed methods. Comparisons are also made between LMVS and PMVS, together with a few well-known semi-supervised clustering algorithms.