Affinity learning via a diffusion process for subspace clustering

作者：

Highlights：

• An effective affinity learning method is proposed.

• This approach utilizes the self-expressive model (L1 norm) as the initial affinities, and re-evaluate them by a diffusion process.

• Learning affinity considering both local geometry and global data manifold.

• Extensive experiments on various data have demonstrated the effectiveness of the proposed method.

摘要

•An effective affinity learning method is proposed.•This approach utilizes the self-expressive model (L1 norm) as the initial affinities, and re-evaluate them by a diffusion process.•Learning affinity considering both local geometry and global data manifold.•Extensive experiments on various data have demonstrated the effectiveness of the proposed method.

论文关键词：Subspace clustering,Diffusion process,Affinity learning

论文评审过程：Received 14 August 2017, Revised 11 May 2018, Accepted 1 July 2018, Available online 2 July 2018, Version of Record 18 July 2018.

论文官网地址：https://doi.org/10.1016/j.patcog.2018.07.002