A Clustering Algorithm Based on Semi-supervised Dimensionality Reduction
Zhang Dao-qiang · Journal of Guangxi Normal University · 2008
Semi-supervised clustering algorithms use a small amount of supervision information in the form of labeled data or pairwise constraints to improve clustering performance.A semi-supervised clustering algorithm based on semi-supervised dimensionality reduction is proposed,the new algorithm contains two subsequent procedures.Firstly,semi-supervised dimensionality reduction is used to project original data into lower dimensional space,and then semisupervised clustering is performed in the reduced space.Because both dimensionality reduction and clustering procedures have taken advantage of the supervision information,the new method can achieve further improvement on clustering performance.Finally,experimental results on a wide range of datasets including UCI machine learning repository,yale and yaleB face databases and text data,validate the effectiveness of the proposed algorithm.