Fair active learning

作者：

Highlights：

• We introduce fair active learning to mitigate bias in limited labeled data problems.

• We Introduce a generic framework that covers different definitions of fairness.

• We Propose a principled manner to balance the fairness–accuracy trade-off.

• The comprehensive experiments confirm the effectiveness of proposed techniques.

摘要

•We introduce fair active learning to mitigate bias in limited labeled data problems.•We Introduce a generic framework that covers different definitions of fairness.•We Propose a principled manner to balance the fairness–accuracy trade-off.•The comprehensive experiments confirm the effectiveness of proposed techniques.

论文关键词：Active learning,Algorithmic fairness,Limited labeled data

论文评审过程：Received 24 August 2021, Revised 15 January 2022, Accepted 23 March 2022, Available online 4 April 2022, Version of Record 16 April 2022.

论文官网地址：https://doi.org/10.1016/j.eswa.2022.116981