Reproduce “Surrogate Losses for Fairness-Aware Classification: Fisher Consistency and Calibration” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Changcheng Song, Yi Yang, Xiangyu Chang, Xiaoling Lu, Yufeng Liu
The authors characterize when smooth surrogate losses preserve decisions in fairness-aware classification, showing that consistency depends on joint derivative conditions for predictive and fairness losses. They derive a calibration inequality linking surrogate-risk error to fairness-aware-risk error. On Give Me Some Credit, an SVM setting reduced equal-opportunity disparity to 0.0015 while measured risk changed from 0.0401 to 0.0399; this result is specific to that study setup.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.