Fair Meta-Learning: Learning How to Learn Fairly

TL;DR

This paper introduces Fair-MAML, a meta-learning approach that enables training fair and accurate models from limited data by incorporating fairness regularization into the MAML algorithm.

Contribution

It adapts the MAML meta-learning algorithm with fairness regularization to efficiently learn fair models from few examples across related tasks.

Findings

Fair-MAML achieves fair and accurate models with minimal data

The approach outperforms relevant baselines in experiments

Meta-learning accelerates fair model training from limited data

Abstract

Data sets for fairness relevant tasks can lack examples or be biased according to a specific label in a sensitive attribute. We demonstrate the usefulness of weight based meta-learning approaches in such situations. For models that can be trained through gradient descent, we demonstrate that there are some parameter configurations that allow models to be optimized from a few number of gradient steps and with minimal data which are both fair and accurate. To learn such weight sets, we adapt the popular MAML algorithm to Fair-MAML by the inclusion of a fairness regularization term. In practice, Fair-MAML allows practitioners to train fair machine learning models from only a few examples when data from related tasks is available. We empirically exhibit the value of this technique by comparing to relevant baselines.