Dual space latent representation learning for unsupervised feature selection
作者：
Highlights：
• This paper proposes a novel feature selection algorithm called DSLRL.
• DSLRL exploits the internal association information of data space and feature space to guide feature selection.
• DSLRL proposes dual space latent representation learning, which characterizes the inherent structure of data space and feature space.
• DSLRL optimizes the low-dimensional latent representation matrix of data space as a pseudo-label matrix to provide clustering indicators.
• DSLRL makes full use of the intrinsic information of dual space to guide the learning of sparse transformation matrix.
• Experimental results on real datasets show the effectiveness of DSLRL.
摘要
•This paper proposes a novel feature selection algorithm called DSLRL.•DSLRL exploits the internal association information of data space and feature space to guide feature selection.•DSLRL proposes dual space latent representation learning, which characterizes the inherent structure of data space and feature space.•DSLRL optimizes the low-dimensional latent representation matrix of data space as a pseudo-label matrix to provide clustering indicators.•DSLRL makes full use of the intrinsic information of dual space to guide the learning of sparse transformation matrix.•Experimental results on real datasets show the effectiveness of DSLRL.
论文关键词：Latent representation learning,Unsupervised feature selection,Dual space,Sparse regression
论文评审过程：Received 27 March 2020, Revised 3 December 2020, Accepted 31 January 2021, Available online 2 February 2021, Version of Record 11 February 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.107873
