AI Paper Atlas

Research map / Learning theory and algorithms / Fairness

FairGRAPE: Fairness-Aware GRAdient Pruning mEthod for Face Attribute Classification

Xiaofeng Lin, Seungbae Kim, Jungseock Joo

ECCV 2022 · Learning theory and algorithms › Fairness · approach: Fairness-constrained classification · 43 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
add group-wise importance and fair node selection to pruningFairGRAPE · FairFace race (99% sparsity)accuracy: 62.3 -> 66.8Improved
add group importance (ablation)FairGRAPE · CelebA attributesaccuracy: 89.31 -> 90.90Improved
add group importance (ablation)FairGRAPE · CelebA attributesstd of accuracy loss: 0.31 -> 0.11Improved
add iterative retraining (ablation)FairGRAPE · CelebA attributesaccuracy: 89.26 -> 90.90Improved
increase pruning step r from 0.1 to 0.2 (ablation)FairGRAPE · CelebA attributesaccuracy: 90.90 -> 90.49Hurt
increase pruning step r to 0.5 (ablation)FairGRAPE · CelebA attributesaccuracy: 90.90 -> 90.34Hurt
use 100% instead of 20% training images for importance (ablation)FairGRAPE · CelebA attributesaccuracy: 90.90 -> 90.66Hurt
compute pruning on minority-race examples only (ablation)GraSP · UTKFace genderstd of accuracy loss: 1.77 -> 2.31Hurt
compute pruning on minority-race examples only (ablation)SNIP · UTKFace genderstd of accuracy loss: 1.62 -> 1.90Hurt
compute pruning on minority-race examples only (ablation)SNIP · UTKFace genderoverall accuracy: 90.95 -> 90.55No clear effect
use K-means clusters instead of race labels as groupsFairGRAPE · FairFace racestd of accuracy loss: 11.3 -> 7.40Improved
apply FairGRAPE instead of GraSP on ImageNet person subtreeFairGRAPE · ImageNet person subtreeaccuracy: 15.36 -> 51.12Improved

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