Global and intrinsic geometric structure embedding for unsupervised feature selection

作者：

Highlights：

• Considering both the loss of information and structure embedding.

• Global and intrinsic geometric structure is preserved by low-rank-sparse graph embedding.

• l2,1/2-Norm on projection matrix help to select more sparse and discriminative features.

• Experiments prove that our GGEFS outperforms other methods.

摘要

•Considering both the loss of information and structure embedding.•Global and intrinsic geometric structure is preserved by low-rank-sparse graph embedding.•l2,1/2-Norm on projection matrix help to select more sparse and discriminative features.•Experiments prove that our GGEFS outperforms other methods.

论文关键词：Sparse preserve projection,Low rank representation,l2,1/2-Matrix norm,Unsupervised feature selection

论文评审过程：Received 14 June 2017, Revised 14 September 2017, Accepted 3 October 2017, Available online 4 October 2017, Version of Record 14 October 2017.

论文官网地址：https://doi.org/10.1016/j.eswa.2017.10.008