Simple and near-optimal algorithms for hidden stratification and multi-group learning

TL;DR

This paper introduces simple, near-optimal algorithms for multi-group learning, addressing hidden stratification and subgroup fairness concerns by analyzing solution structures.

Contribution

It provides a theoretical analysis of multi-group learning solutions and proposes algorithms that are both simple and near-optimal.

Findings

Algorithms achieve near-optimal performance

Addresses hidden stratification issues

Enhances subgroup fairness in learning models

Abstract

Multi-group agnostic learning is a formal learning criterion that is concerned with the conditional risks of predictors within subgroups of a population. The criterion addresses recent practical concerns such as subgroup fairness and hidden stratification. This paper studies the structure of solutions to the multi-group learning problem, and provides simple and near-optimal algorithms for the learning problem.