AI Paper Atlas

Research map / Learning theory and algorithms / Fairness

A Unified Approach to Fair Online Learning via Blackwell Approachability

Evgenii E Chzhen, Christophe Giraud, Gilles Stoltz

NeurIPS 2021 · Learning theory and algorithms › Fairness · approach: Fairness-constrained classification · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use Blackwell approachability with sequential estimates of the unknown context distributionFair online learning with stochastic contexts · General approachability settingdistance to target setImproved
Combine vanilla calibration with demographic parityOnline calibration · Vanilla calibration with demographic paritycalibration error and demographic parity violation: 1/NImproved
Require group-wise no-regret when the sensitive context is irrelevant or identifiable from the non-sensitive contextOnline no-regret · Group-wise no-regretgroup-wise regretImproved
Require group-wise no-regret with overlapping group-conditional context supports in the counterexamplesOnline no-regret · Group-wise no-regretgroup-wise regretHurt
Allow a demographic-parity violation level of δτ and optimize the calibration toleranceGroup-wise calibration · Nature observes sensitive contextcalibration error: 1 - τ · TV(Q⁰, Q¹) ≤ ε*(δτ) ≤ 1 - τ · TV(Q⁰, Q¹) + 1/NMixed
Allow a demographic-parity violation level of δτ and optimize the calibration toleranceGroup-wise calibration · Nature does not observe sensitive contextcalibration error: (1 - τ) · TV(Q⁰, Q¹) ≤ ε*(δτ) ≤ (1 - τ) · TV(Q⁰, Q¹) + 1/NMixed
Set τ = 1 when Nature does not observe the sensitive contextGroup-wise calibration · Group-wise calibration with demographic paritycalibration error: 1/NImproved
Require both calibration error and demographic-parity violation to be at most 1/N when Nature observes the sensitive contextGroup-wise calibration · Group-wise calibration with demographic paritycalibration error: ε = 1/NHurt

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy