Matrix completion by least-square, low-rank, and sparse self-representations

作者：

Highlights：

• In this paper, the problem of missing value is studied. We propose a novel framework of matrix completion, which is based on self-representation.

• Specifically, least-square, low-rank, and sparse self-representations based matrix completion algorithms are proposed.

• The proposed algorithms are able to recover the missing values when the data are from multiple subspaces, even if the matrices are of high-rank.

• The proposed algorithms are compared with state-of-the-art algorithms in the tasks of synthetic matrix completion, image inpainting, and collaborative filtering.

摘要

•In this paper, the problem of missing value is studied. We propose a novel framework of matrix completion, which is based on self-representation.•Specifically, least-square, low-rank, and sparse self-representations based matrix completion algorithms are proposed.•The proposed algorithms are able to recover the missing values when the data are from multiple subspaces, even if the matrices are of high-rank.•The proposed algorithms are compared with state-of-the-art algorithms in the tasks of synthetic matrix completion, image inpainting, and collaborative filtering.

论文关键词：Matrix completion,Missing value,Low-rank and sparse representations,Image inpainting,Collaborative filtering

论文评审过程：Received 29 December 2016, Revised 4 May 2017, Accepted 13 May 2017, Available online 15 May 2017, Version of Record 21 June 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.05.013