An improved density peak clustering algorithm guided by pseudo labels

作者：

Highlights：

• We propose a new clustering framework without the need of parameter adjustment by using pseudo labels.

• We try to maximize mutual information in order to represent the relationship between pseudo labels and parameters.

• The PLOPC outperforms eight parametric clustering algorithms with the best parameter configuration and three adaptive clustering algorithms in most cases.

摘要

•We propose a new clustering framework without the need of parameter adjustment by using pseudo labels.•We try to maximize mutual information in order to represent the relationship between pseudo labels and parameters.•The PLOPC outperforms eight parametric clustering algorithms with the best parameter configuration and three adaptive clustering algorithms in most cases.

论文关键词：Density peak clustering,Pseudo labels,Maximizing mutual information

论文评审过程：Received 18 April 2022, Revised 2 July 2022, Accepted 2 July 2022, Available online 8 July 2022, Version of Record 16 July 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.109374