Refining a k-nearest neighbor graph for a computationally efficient spectral clustering

作者：

Highlights：

• A refined k-nearest neighbor graph to reduce memory needed for spectral clustering.

• The method is deterministic and performs consistently on independent executions.

• The method scores higher than ASC methods in terms of clustering evaluation metrics.

摘要

•A refined k-nearest neighbor graph to reduce memory needed for spectral clustering.•The method is deterministic and performs consistently on independent executions.•The method scores higher than ASC methods in terms of clustering evaluation metrics.

论文关键词：Spectral clustering,Approximate spectral clustering,k-nearest neighbor graph,Local scale similarity

论文评审过程：Received 14 June 2019, Revised 22 July 2020, Accepted 31 January 2021, Available online 6 February 2021, Version of Record 13 February 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.107869