Technical Challenges for Training Fair Neural Networks

Abstract

As machine learning algorithms have been widely deployed across applications, many concerns have been raised over the fairness of their predictions, especially in high stakes settings (such as facial recognition and medical imaging). To respond to these concerns, the community has proposed and formalized various notions of fairness as well as methods for rectifying unfair behavior. While fairness constraints have been studied extensively for classical models, the effectiveness of methods for imposing fairness on deep neural networks is unclear. In this paper, we observe that these large models overfit to fairness objectives, and produce a range of unintended and undesirable consequences. We conduct our experiments on both facial recognition and automated medical diagnosis datasets using state-of-the-art architectures.

References in corpus (7)

• Equality of Opportunity in Supervised Learning
• Fairness Testing: Testing Software for Discrimination
• Data Decisions and Theoretical Implications when Adversarially Learning Fair Representations
• Minimax Pareto Fairness: A Multi Objective Perspective
• On Adversarial Bias and the Robustness of Fair Machine Learning
• On the Long-term Impact of Algorithmic Decision Policies: Effort Unfairness and Feature Segregation through Social Learning
• LowKey: Leveraging Adversarial Attacks to Protect Social Media Users from Facial Recognition