A hybrid safe semi-supervised learning method

作者：

Highlights：

• A hybrid S3L method is proposed to inherit the merits of existing two approaches.

• Safely exploit the risky unlabeled instances and select the high-quality graphs.

• An alternating iterative strategy is introduced to solve the optimization problem.

• Our algorithm achieves the promising performance and expected goal.

摘要

•A hybrid S3L method is proposed to inherit the merits of existing two approaches.•Safely exploit the risky unlabeled instances and select the high-quality graphs.•An alternating iterative strategy is introduced to solve the optimization problem.•Our algorithm achieves the promising performance and expected goal.

论文关键词：Semi-supervised learning,Risk degree,Graph quality,Laplacian regularized least squares

论文评审过程：Received 24 November 2018, Revised 20 November 2019, Accepted 5 February 2020, Available online 6 February 2020, Version of Record 15 February 2020.

论文官网地址：https://doi.org/10.1016/j.eswa.2020.113295