AI Paper Atlas

Research map / Learning theory and algorithms / Fairness

Are Two Heads the Same as One? Identifying Disparate Treatment in Fair Neural Networks

Michael Lohaus, Matthäus Kleindessner, Krishnaram Kenthapadi, Francesco Locatello, Chris Russell

NeurIPS 2022 · Learning theory and algorithms › Fairness · approach: Fairness-constrained classification · 4 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 two-headed network with merged heads instead of Lipton per-group thresholdstwo-headed approach · CelebA Attractive, 80% disparity reductionaccuracy: 0.7719 -> 0.7698No clear effect
use two-headed approach instead of fairness regularizerregularizer R_DP · CelebA, 50% and 80% disparity reductionachieving required fairnessImproved
use two-headed approach instead of Massaging preprocessingMassaging · CelebA, 80% disparity reductionachieving required fairnessImproved
increase fairness parameter lambdafairness regularizer R_DP (ResNet50) · CelebA, 72 target-protected pairsprotected attribute recovery accuracy from last layer: 71 out of 72Improved
increase fairness parameter lambdafairness regularizer R_DP · CelebA, one pairprotected attribute recovery (Kendall tau): p=0.64No clear effect
replace the group output g(x) with the median output of the other groupfairest regularized model · CelebAproportion of changed predictions: around $30\%$Mixed
replace the group output g(x) in MassagingMassaging · CelebAproportion of changed predictions: up to $35\%$Mixed
use regularizer instead of two-headed approach to reach near-perfect fairnessregularizer · CelebA Attractive vs MaleDDP reached: saturates around $-0.1$Hurt
use abs-DP regularizer instead of R_DP (ablation)regularizer R_DP^abs · CelebA Smiling, MobileNetV3-Small, 50% reductionaccuracy: 0.4997Hurt
swap ResNet50 for MobileNetV3-Smalltwo-headed approach · CelebA Attractive, 50% reductionaccuracy: 0.8021 -> 0.8023No clear effect

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