Maximum likelihood based pairwise clustering
Xiaobin Li, Sanyang Liu, Liu Mige, Zheng Tian · 2011
This paper presents a novel pairwise clustering approach. We pose the problem as a question of parameter estimation and show the pairwise indicator variables can be estimated by using the maximum likelihood estimate (MLE) method. Based on this, a two-level clustering algorithm is developed: the grouping graph is first condensed by using the MLE results and then the k-means clustering method is applied directly to the condensed graph of much small size. We have applied our algorithm to a number of artificial and real-world data sets, and found the results to be very encouraging.