OxEnsemble: Fair Ensembles for Low-Data Classification
It addresses fairness in low-data regimes for domains like medical imaging, where false negatives can be fatal, but it appears incremental as it builds on ensemble methods with fairness constraints.
The paper tackles fair classification in low-data, unbalanced settings like medical imaging, proposing OxEnsemble to train ensembles with fairness constraints, resulting in more consistent outcomes and stronger fairness-accuracy trade-offs than existing methods across datasets.
We address the problem of fair classification in settings where data is scarce and unbalanced across demographic groups. Such low-data regimes are common in domains like medical imaging, where false negatives can have fatal consequences. We propose a novel approach \emph{OxEnsemble} for efficiently training ensembles and enforcing fairness in these low-data regimes. Unlike other approaches, we aggregate predictions across ensemble members, each trained to satisfy fairness constraints. By construction, \emph{OxEnsemble} is both data-efficient, carefully reusing held-out data to enforce fairness reliably, and compute-efficient, requiring little more compute than used to fine-tune or evaluate an existing model. We validate this approach with new theoretical guarantees. Experimentally, our approach yields more consistent outcomes and stronger fairness-accuracy trade-offs than existing methods across multiple challenging medical imaging classification datasets.