Penta-Training: Clustering Ensembles with Bootstrapping of Constraints

In this paper we combine clustering ensembles and semisupervised clustering to address the ill-posed nature of clustering. We introduce a mechanism which leverages the ensemble framework to bootstrap informative constraints directly from the data and from the various clusterings, without intervention from the user. Our approach is well suited for problems where the information available from an external source is very limited. We demonstrate the effectiveness of our proposed technique with experiments using real datasets and other state-of-the-art semi-supervised techniques.