Information maximization clustering via multi-view self-labelling

作者：

Highlights：

• Modifying and leveraging the mutual information to cluster the image data through an over-clustering distribution.

• Converting grouping-based self-supervised methods into multi-functional frameworks.

• The clustering is achieved jointly in a single-phase training process without increasing the training stages or hyper-parameters.

摘要

•Modifying and leveraging the mutual information to cluster the image data through an over-clustering distribution.•Converting grouping-based self-supervised methods into multi-functional frameworks.•The clustering is achieved jointly in a single-phase training process without increasing the training stages or hyper-parameters.

论文关键词：Deep neural models,Mutual information maximization,Unsupervised learning,Self-supervised learning,Image clustering

论文评审过程：Received 2 December 2021, Revised 25 March 2022, Accepted 10 May 2022, Available online 23 May 2022, Version of Record 8 June 2022.

论文官网地址：https://doi.org/10.1016/j.knosys.2022.109042