Extended Semi-supervised Matrix Factorization for Clustering

Abstract

In this paper, we extend the Penalized Matrix Factorization (PMF) algorithm for semi-supervised clustering. The definition of may-link constraints are introduced and obtained based on must-link constraints and cluster structure. We derive the Extended PMF (EPMF) model by incorporating the may-link constraints inside the original PMF decomposition. Extensive experimental evaluations are performed on the SECTOR data set. The experimental results show the effectiveness of the extended PMF.