Active semi-supervised affinity propagation clustering algorithm based on pair-wise constraints
Qi Lei, Huiping Yu, Min Wu · 2014
Pair-wise constraints are widely used in semi-supervised clustering to aid unsupervised learning, but traditional semi-supervised clustering algorithm lacks the ability to find the useful constraint information. This paper presents a semi-supervised affinity propagation(AP) clustering algorithm based on active learning, which can select informative pair-wise constraints to find constraint information that cannot be noticed by the clustering algorithm easily. The constraint information obtained with the active learning method is used to adjust the similarity matrix in the AP clustering algorithm and make it semi-supervised with side information. We compare our method with the AP clustering algorithm and K-means algorithm, both with constraints selected randomly. Experimental results on the UCI Machine Learning Repository indicate that the new clustering algorithm proposed in this paper can improve the clustering performance significantly.