AFed: Algorithmic Fair Federated Learning
Huiqiang Chen, Tianqing Zhu, Wanlei Zhou, and Wei Zhao · 2025
Read AFed: Algorithmic Fair Federated Learning by Huiqiang Chen, Tianqing Zhu, Wanlei Zhou, and Wei Zhao (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-02732.txt.
Description
Huiqiang Chen, Tianqing Zhu, Wanlei Zhou, and Wei Zhao tackle fairness inside federated learning, where phones and institutions train a shared model without pooling their data. Keeping data local protects privacy but blocks standard debiasing tools, which need central access to sensitive attributes. Skewed local data also makes federated models more prone to group unfairness than centralized ones.
Their framework, AFed, works around the access problem by learning what the global data distribution looks like. One variant trains a conditional generator on the server; the stronger variant, AFed-GAN, trains a conditional adversarial generator on the client side. Clients then pad their own data with generated samples that help wash out bias. The authors back the design with theory and test it on several real-world datasets, reporting substantial fairness gains over baseline methods while raw data never leaves its owners.
Catalog record
| CREATOR | Huiqiang Chen, Tianqing Zhu, Wanlei Zhou, and Wei Zhao |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 66 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-02732