A multi-kernel method of measuring adaptive similarity for spectral clustering

作者：

Highlights：

• Multiple kernels extract good global structure of data.

• Adoptive and optimal neighbors for data point based on local structure is learnt.

• Choice of appropriate kernels for specific tasks is simplified.

• Learns similarity matrix and label matrix simultaneously.

摘要

•Multiple kernels extract good global structure of data.•Adoptive and optimal neighbors for data point based on local structure is learnt.•Choice of appropriate kernels for specific tasks is simplified.•Learns similarity matrix and label matrix simultaneously.

论文关键词：Adaptive neighbors,Multi-kernel,Reproducing kernel Hilbert space,Similarity measure,Spectral clustering

论文评审过程：Received 26 April 2019, Revised 10 February 2020, Accepted 13 May 2020, Available online 16 May 2020, Version of Record 30 May 2020.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.113570