Advances in Fair Machine Learning

The field of machine learning is moving towards a greater emphasis on fairness and transparency, with a focus on developing methods that can provide strong guarantees on the fairness of the learned models. Recent work has explored the use of Lagrangian duality and PAC-Bayes to address constrained learning problems and provide generalization guarantees. There is also a growing recognition of the importance of high-quality datasets and robust evaluation methodologies in ensuring the reliability and reproducibility of fair machine learning research. Notable papers in this area include: AL-CoLe, which establishes strong duality results for Augmented Lagrangian methods in non-convex settings. Fair Representation Learning with Controllable High Confidence Guarantees via Adversarial Inference, which proposes a framework for learning representations that achieve high-confidence fairness guarantees. Bias Begins with Data: The FairGround Corpus, which presents a unified framework and dataset corpus for advancing reproducible research in fair machine learning classification.

Sources

AL-CoLe: Augmented Lagrangian for Constrained Learning

Fair Representation Learning with Controllable High Confidence Guarantees via Adversarial Inference

Bias Begins with Data: The FairGround Corpus for Robust and Reproducible Research on Algorithmic Fairness

A Framework for Bounding Deterministic Risk with PAC-Bayes: Applications to Majority Votes

MLPrE -- A tool for preprocessing and exploratory data analysis prior to machine learning model construction

Built with on top of