AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Learning Debiased Classifier with Biased Committee

Nayeong Kim, Sehyun Hwang, Sungsoo Ahn, Jaesik Park, Suha Kwak

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Learning from unlabeled and positive-unlabeled data · 13 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
up-weight samples misclassified by a single biased classifier (vs plain ERM) (ablation)SSL+ERM on CelebA HairColor · CelebA HairColorWorst-group accuracy: 38.5 -> 64.1Improved
use a committee of 40 biased classifiers instead of a single one (ablation)LWBC · CelebA HairColorWorst-group accuracy: 64.1 -> 81.3Improved
add knowledge distillation from main classifier to the committee (ablation)LWBC · CelebA HairColorWorst-group accuracy: 81.3 -> 85.5Improved
add knowledge distillation to a single-classifier variant (ablation)LWBC single member · CelebA HairColorWorst-group accuracy: 72.7 -> 77.9Improved
train committee members on bootstrapped random subsets instead of entire data (ablation)LWBC · CelebA HairColorWorst-group accuracy: 78.0 -> 81.3Improved
replace JTT weighting with LWBC weight function (ablation)SSL single classifier · CelebA HairColorWorst-group accuracy: 57.7 -> 64.1Improved
use self-supervised features (vs none) (ablation)LWBC · CelebA HairColorWorst-group accuracy: 80.8 -> 85.5Improved
bootstrapped committee vs deep ensemble (ablation)LWBC · CelebA HairColorWorst-group accuracy: 77.2 -> 85.5Improved
increase committee size beyond 40 (ablation)LWBC · CelebA HairColorWorst-group accuracyNo clear effect
self-supervised representation instead of supervised ERMERM · CelebA HairColorConflicting accuracy: 52.52 -> 66.79Improved
add LWBC on top of SSL+ERMSSL+ERM · BARConflicting accuracy: 60.88 -> 62.03Improved
add LWBC on top of SSL+ERMSSL+ERM · ImageNet-9Unbiased accuracy: 93.18 -> 93.04No 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