conftrace_

2022 ICML ICML 2022

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

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.

Authors

Christopher J Tosh, Daniel Hsu

Topics

Machine Learning > Core Methods > Classification Machine Learning > Optimization & Theory > Learning Theory Machine Learning > Application Areas > Fairness

Keywords

agnostic learning subgroup fairness multi-group learning hidden stratification conditional risk near-optimal algorithm

More from ICML 2022

All ICML 2022 papers →

Failure and success of the spectral bias prediction for Laplace Kernel Ridge Regression: the case of low-dimensional data

Quantifying and Learning Linear Symmetry-Based Disentanglement

A Temporal-Difference Approach to Policy Gradient Estimation

Design-Bench: Benchmarks for Data-Driven Offline Model-Based Optimization

AnyMorph: Learning Transferable Polices By Inferring Agent Morphology

Detecting Adversarial Examples Is (Nearly) As Hard As Classifying Them