Fair Representation Learning with Unreliable Labels

Yixuan Zhang, Feng Zhou, Zhidong Li, Yang Wang, Fang Chen
Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, PMLR 206:4655-4667, 2023.

Abstract

In learning with fairness, for every instance, its label can be randomly flipped to another class due to the practitioner’s prejudice, namely, label bias. The existing well-studied fair representation learning methods focus on removing the dependency between the sensitive factors and the input data, but do not address how the representations retain useful information when the labels are unreliable. In fact, we find that the learned representations become random or degenerated when the instance is contaminated by label bias. To alleviate this issue, we investigate the problem of learning fair representations that are independent of the sensitive factors while retaining the task-relevant information given only access to unreliable labels. Our model disentangles the dependency between fair representations and sensitive factors in the latent space. To remove the reliance between the labels and sensitive factors, we incorporate an additional penalty based on mutual information. The learned purged fair representations can then be used in any downstream processing. We demonstrate the superiority of our method over previous works through multiple experiments on both synthetic and real-world datasets.

Cite this Paper

BibTeX
@InProceedings{pmlr-v206-zhang23g, title = {Fair Representation Learning with Unreliable Labels}, author = {Zhang, Yixuan and Zhou, Feng and Li, Zhidong and Wang, Yang and Chen, Fang}, booktitle = {Proceedings of The 26th International Conference on Artificial Intelligence and Statistics}, pages = {4655--4667}, year = {2023}, editor = {Ruiz, Francisco and Dy, Jennifer and van de Meent, Jan-Willem}, volume = {206}, series = {Proceedings of Machine Learning Research}, month = {25--27 Apr}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v206/zhang23g/zhang23g.pdf}, url = {https://proceedings.mlr.press/v206/zhang23g.html}, abstract = {In learning with fairness, for every instance, its label can be randomly flipped to another class due to the practitioner’s prejudice, namely, label bias. The existing well-studied fair representation learning methods focus on removing the dependency between the sensitive factors and the input data, but do not address how the representations retain useful information when the labels are unreliable. In fact, we find that the learned representations become random or degenerated when the instance is contaminated by label bias. To alleviate this issue, we investigate the problem of learning fair representations that are independent of the sensitive factors while retaining the task-relevant information given only access to unreliable labels. Our model disentangles the dependency between fair representations and sensitive factors in the latent space. To remove the reliance between the labels and sensitive factors, we incorporate an additional penalty based on mutual information. The learned purged fair representations can then be used in any downstream processing. We demonstrate the superiority of our method over previous works through multiple experiments on both synthetic and real-world datasets.} }
Endnote
%0 Conference Paper %T Fair Representation Learning with Unreliable Labels %A Yixuan Zhang %A Feng Zhou %A Zhidong Li %A Yang Wang %A Fang Chen %B Proceedings of The 26th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2023 %E Francisco Ruiz %E Jennifer Dy %E Jan-Willem van de Meent %F pmlr-v206-zhang23g %I PMLR %P 4655--4667 %U https://proceedings.mlr.press/v206/zhang23g.html %V 206 %X In learning with fairness, for every instance, its label can be randomly flipped to another class due to the practitioner’s prejudice, namely, label bias. The existing well-studied fair representation learning methods focus on removing the dependency between the sensitive factors and the input data, but do not address how the representations retain useful information when the labels are unreliable. In fact, we find that the learned representations become random or degenerated when the instance is contaminated by label bias. To alleviate this issue, we investigate the problem of learning fair representations that are independent of the sensitive factors while retaining the task-relevant information given only access to unreliable labels. Our model disentangles the dependency between fair representations and sensitive factors in the latent space. To remove the reliance between the labels and sensitive factors, we incorporate an additional penalty based on mutual information. The learned purged fair representations can then be used in any downstream processing. We demonstrate the superiority of our method over previous works through multiple experiments on both synthetic and real-world datasets.
APA
Zhang, Y., Zhou, F., Li, Z., Wang, Y. & Chen, F.. (2023). Fair Representation Learning with Unreliable Labels. Proceedings of The 26th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 206:4655-4667 Available from https://proceedings.mlr.press/v206/zhang23g.html.

Related Material