Summary of Fairness Without Demographics in Human-centered Federated Learning, by Shaily Roy et al.
Fairness Without Demographics in Human-Centered Federated Learning
by Shaily Roy, Harshit Sharma, Asif Salekin
First submitted to arxiv on: 30 Apr 2024
Categories
- Main: Machine Learning (cs.LG)
- Secondary: Artificial Intelligence (cs.AI); Distributed, Parallel, and Cluster Computing (cs.DC)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary Federated learning (FL) is a privacy-preserving technique for collaborative model training. However, ensuring fairness in these systems remains a significant research gap. Current approaches require knowledge of sensitive attributes, which contradicts FL’s privacy principles. Our novel approach, inspired by “Fairness without Demographics” in machine learning, achieves fairness without needing sensitive attribute information. We minimize the top eigenvalue of the Hessian matrix during training to ensure equitable loss landscapes across FL participants. Our method also introduces a new FL aggregation scheme that promotes participating models based on error rates and loss landscape curvature attributes, fostering fairness. This work represents the first approach to attaining “Fairness without Demographics” in human-centered FL. We demonstrate effectiveness in balancing fairness and efficacy through comprehensive evaluation across various real-world applications, FL setups, and scenarios involving single and multiple bias-inducing factors. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Imagine a way for people or devices to work together to create better artificial intelligence models without sharing their personal data. This is called federated learning (FL). But right now, there’s a problem with making sure everyone gets treated fairly in these systems. Our solution doesn’t need to know what makes each person different to ensure fairness. We use a special technique that helps make the loss landscape (how well the model performs) fair for all participants. This is the first time this approach has been used in human-centered FL. By testing our method, we showed it can balance fairness and performance across many real-world scenarios. |
Keywords
» Artificial intelligence » Federated learning » Machine learning
