AI Paper Atlas

Research map / Label-efficient and robust learning / OOD and anomaly detection

Not All Out-of-Distribution Data Are Harmful to Open-Set Active Learning

Yang Yang, Yuxuan Zhang, XIN SONG, Yi Xu

NeurIPS 2023 · Label-efficient and robust learning › OOD and anomaly detection · approach: Provable and feature-based OOD detection · 3 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
replace mu=0.8 weighting with mu=0.2 (ablation)PAL · CIFAR-100 20% IDaccuracy: 69.4 -> 66.2Hurt
use mu=1 (informativeness only dominance) instead of 0.8 (ablation)PAL · CIFAR-100 20% IDaccuracy: 69.4 -> 67.6Hurt
lower mu from 0.8 to 0.2 for precision (ablation)PAL · CIFAR-100 20% IDprecision: 25.4 -> 32.4Improved
use mu=0.8 versus 1 for OOD detection recall (ablation)PAL · CIFAR-100 20% IDrecall: 69.7 -> 73.8Improved
replace meta-weight with Coreset (PAL_Core) (ablation)PAL · CIFAR-100 20% IDaccuracy: 69.4 -> 65.9Hurt
replace informativeness criterion with MSP (PAL_MSP) (ablation)PAL · CIFAR-100 20% IDaccuracy: 69.4 -> 64.7Hurt
replace informativeness criterion with energy score (PAL_energy) (ablation)PAL · CIFAR-100 20% IDaccuracy: 69.4 -> 66.2Hurt
replace informativeness criterion with MSP (PAL_MSP) (ablation)PAL · CIFAR-10 20% IDaccuracy: 98.7 -> 98.0Hurt
remove representativeness criterion or informativeness criterion (w/o omega, w/o SID) (ablation)PAL · CIFAR-10, CIFAR-100, Tiny-Imagenet 20% IDaccuracyHurt
actively query pseudo-OOD instances (PAL) instead of only pseudo-ID (LfOSA)LfOSA · Tiny-Imagenet 20% IDaccuracy: 47.9 -> 52.1Improved
actively query pseudo-OOD instances (PAL) instead of only pseudo-ID (LfOSA)LfOSA · CIFAR-100 20% IDaccuracy: 66.1 -> 69.4Improved
actively query pseudo-OOD instances (PAL) instead of only pseudo-ID (LfOSA)LfOSA · CIFAR-10 20% ID, round 1accuracy: 84.2 -> 91.1Improved

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