Metric-Fair Active Learning

Jie Shen, Nan Cui, Jing Wang
Proceedings of the 39th International Conference on Machine Learning, PMLR 162:19809-19826, 2022.

Abstract

Active learning has become a prevalent technique for designing label-efficient algorithms, where the central principle is to only query and fit “informative” labeled instances. It is, however, known that an active learning algorithm may incur unfairness due to such instance selection procedure. In this paper, we henceforth study metric-fair active learning of homogeneous halfspaces, and show that under the distribution-dependent PAC learning model, fairness and label efficiency can be achieved simultaneously. We further propose two extensions of our main results: 1) we show that it is possible to make the algorithm robust to the adversarial noise – one of the most challenging noise models in learning theory; and 2) it is possible to significantly improve the label complexity when the underlying halfspace is sparse.

Cite this Paper

BibTeX
@InProceedings{pmlr-v162-shen22b, title = {Metric-Fair Active Learning}, author = {Shen, Jie and Cui, Nan and Wang, Jing}, booktitle = {Proceedings of the 39th International Conference on Machine Learning}, pages = {19809--19826}, year = {2022}, editor = {Chaudhuri, Kamalika and Jegelka, Stefanie and Song, Le and Szepesvari, Csaba and Niu, Gang and Sabato, Sivan}, volume = {162}, series = {Proceedings of Machine Learning Research}, month = {17--23 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v162/shen22b/shen22b.pdf}, url = {https://proceedings.mlr.press/v162/shen22b.html}, abstract = {Active learning has become a prevalent technique for designing label-efficient algorithms, where the central principle is to only query and fit “informative” labeled instances. It is, however, known that an active learning algorithm may incur unfairness due to such instance selection procedure. In this paper, we henceforth study metric-fair active learning of homogeneous halfspaces, and show that under the distribution-dependent PAC learning model, fairness and label efficiency can be achieved simultaneously. We further propose two extensions of our main results: 1) we show that it is possible to make the algorithm robust to the adversarial noise – one of the most challenging noise models in learning theory; and 2) it is possible to significantly improve the label complexity when the underlying halfspace is sparse.} }
Endnote
%0 Conference Paper %T Metric-Fair Active Learning %A Jie Shen %A Nan Cui %A Jing Wang %B Proceedings of the 39th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2022 %E Kamalika Chaudhuri %E Stefanie Jegelka %E Le Song %E Csaba Szepesvari %E Gang Niu %E Sivan Sabato %F pmlr-v162-shen22b %I PMLR %P 19809--19826 %U https://proceedings.mlr.press/v162/shen22b.html %V 162 %X Active learning has become a prevalent technique for designing label-efficient algorithms, where the central principle is to only query and fit “informative” labeled instances. It is, however, known that an active learning algorithm may incur unfairness due to such instance selection procedure. In this paper, we henceforth study metric-fair active learning of homogeneous halfspaces, and show that under the distribution-dependent PAC learning model, fairness and label efficiency can be achieved simultaneously. We further propose two extensions of our main results: 1) we show that it is possible to make the algorithm robust to the adversarial noise – one of the most challenging noise models in learning theory; and 2) it is possible to significantly improve the label complexity when the underlying halfspace is sparse.
APA
Shen, J., Cui, N. & Wang, J.. (2022). Metric-Fair Active Learning. Proceedings of the 39th International Conference on Machine Learning, in Proceedings of Machine Learning Research 162:19809-19826 Available from https://proceedings.mlr.press/v162/shen22b.html.

Related Material