Welcome to EquiPy’s documentation!#
EquiPy is a library implementing sequential fairness on the predicted outputs of Machine Learning models, when dealing with multiple sensitive attributes. The library contains a set of classes that implement the methodology and a module for visualizations.
Under the hood, we use a post-processing method to progressively achieve fairness accross a set of sensitive features by leveraging multi-marginal Wasserstein barycenters, which extends the standard notion of Strong Demographic Parity to the case with multiple sensitive characteristics. This approach seamlessly extends to approximate fairness, enveloping a framework accommodating the trade-off between performance and unfairness. You can find the technical details in the technical paper https://arxiv.org/abs/2309.06627 (forthcoming at AAAI 2024).
Contents: