Learning a mixture model for clustering with the completed likelihood minimum message length criterion

作者：

Highlights：

• Propose the CL-MML model selection criterion for cluster number selection.

• CL-MML is robust against the overlap among data clusters.

• CL-MML outperforms the existing counterparts if a prior knowledge of the data distribution is not available.

• A variant of EM algorithm is presented to overcome the deficiency of the standard EM.

摘要

Highlights•Propose the CL-MML model selection criterion for cluster number selection.•CL-MML is robust against the overlap among data clusters.•CL-MML outperforms the existing counterparts if a prior knowledge of the data distribution is not available.•A variant of EM algorithm is presented to overcome the deficiency of the standard EM.

论文关键词：Completed likelihood,Minimum message length,Model selection,Clustering,Finite mixture model

论文评审过程：Received 18 September 2012, Revised 19 September 2013, Accepted 30 September 2013, Available online 9 October 2013.

论文官网地址：https://doi.org/10.1016/j.patcog.2013.09.036