Reproduce “Practical Estimation of the Bayes Optimal Fairness-Accuracy Tradeoff with Soft Labels” 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
Mohit Sharma, Okan Koç, Amit Deshpande, Takashi Ishida, Gang Niu, Rajiv Ratn Shah, Masashi Sugiyama
The authors estimate a reference frontier of classification error across demographic-parity or equal-opportunity disparities from group membership and positive-class probability scores, without examples or training a predictor. They estimate group-specific thresholds and measure error and disparity on separate scores. In synthetic comparisons, estimates generally have lower error than compared methods at similar disparity; this is not an unconditional optimum.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.