AI Paper Atlas

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

OOD-MAML: Meta-Learning for Few-Shot Out-of-Distribution Detection and Classification

Taewon Jeong, Heeyoung Kim

NeurIPS 2020 · Label-efficient and robust learning › OOD and anomaly detection · approach: Spurious-aware OOD detection · 33 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 learned fake (adversarial OOD) samples to (N+1)-class MAML (ablation)(N+1)-MAML · Omniglotdetection accuracy (alpha=95): 0.9142 -> 0.9524Improved
add learned fake samples to (N+1)-class MAML (ablation)(N+1)-MAML · CIFAR-FSTNR (alpha=95): 0.1051 -> 0.2017Improved
add learned fake samples to (N+1)-class MAML (ablation)(N+1)-MAML · miniImageNetTNR (alpha=95): 0.0749 -> 0.1009Improved
use OOD-MAML one-class sub-tasks instead of (N+1)-MAML* with fake samples (ablation)OOD-MAML · miniImageNetTNR (alpha=95): 0.1009 -> 0.6770Improved
use OOD-MAML instead of ODIN on MAMLfew-shot OOD detection · CIFAR-FSTNR (alpha=95): 0.2311 -> 0.5492Improved
use task-specific adapted fake samples (Eq. 2) in addition to task-agnostic fake samples (ablation)OOD-MAML · CIFAR-FS feature visualizationdecision boundary tightness / detectionImproved
increase number of fake samples M from 3 to 5 (ablation)OOD-MAML · Omniglotdetection accuracy (alpha=95): 0.9712 -> 0.9701No clear effect
replace MAML with OOD-MAML for classification5-shot 5-way classification · miniImageNetclassification accuracy: 0.5926 -> 0.6044No clear effect
change test-time number of classes N from 5 to 7 without retrainingOOD-MAML · CIFAR-FSclassification accuracy: 0.7158 -> 0.6964Hurt

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