AI Paper Atlas

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

Distributionally Robust Optimization with Probabilistic Group

Soumya Suvra Ghosal, Yixuan Li

AAAI 2023 · Label-efficient and robust learning › Recognition and long-tail · approach: Learning from unlabeled and positive-unlabeled data · 3 citations (OpenAlex)

Read the paper

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 probabilistic group membership rather than hard group labels (ablation)PG-DRO versus G-DRO · Waterbirds and CelebAworst-group accuracy: Waterbirds: 88.0 -> 91.0; CelebA: 88.6 -> 90.0Improved
use probabilistic group membership rather than hard group labels with 5% validation labels (ablation)PG-DRO versus G-DRO · Waterbirds and CelebAworst-group accuracy: Waterbirds: 87.5 -> 89.2; CelebA: 87.5 -> 89.4Improved
use probabilistic group membership with semi-supervised pseudo-labeling (ablation)PG-DRO versus G-DRO · Waterbirdsworst-group accuracy: 89.0 -> 91.0Improved
use probabilistic group membership with semi-supervised pseudo-labeling (ablation)PG-DRO versus G-DRO · CelebAworst-group accuracy: 89.8 -> 89.8No clear effect
use probabilistic group membership rather than hard group labels with zero-shot pseudo-labeling (ablation)PG-DRO versus G-DRO · Waterbirds and CelebAworst-group accuracy: Waterbirds: 87.6 -> 89.4; CelebA: 87.8 -> 88.8Improved
use CLIP zero-shot group probabilities rather than supervised pseudo-labeling (ablation)PG-DRO · Waterbirds and CelebAworst-group accuracy: Waterbirds: 89.2 -> 89.4; CelebA: 89.4 -> 88.8Mixed
tune the generalization adjustment parameter to C=2 rather than use C=0 (ablation)PG-DRO · Waterbirdsworst-group accuracy: +5.1%Improved
reduce group-annotated validation data to 5% (ablation)PG-DRO · Waterbirds and CelebAworst-group accuracy: Waterbirds: 91.0 -> 89.2; CelebA: 90.0 -> 89.4Mixed
use PG-DRO rather than SSAimage classification · Waterbirds and CelebAworst-group accuracy: Waterbirds: 89.0 -> 91.0; CelebA: 89.8 -> 90.0Mixed
use PG-DRO rather than SSAnatural language processing · MultiNLI and CivilComments-WILDSworst-group accuracy: MultiNLI: 76.6 -> 78.9; CivilComments-WILDS: 69.9 -> 73.6Improved
use PG-DRO rather than G-DROrobust training with group probabilities · CelebAworst-group accuracy: 88.7 -> 90.0Improved

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