AI Paper Atlas

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

Two Sides of Meta-Learning Evaluation: In vs. Out of Distribution

Amrith Setlur, Oscar Li, Virginia Smith

NeurIPS 2021 · Label-efficient and robust learning › OOD and anomaly detection · approach: Out-of-distribution generalization theory · 1 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
Train PN with more ways than test (64-way vs 5-way) (ablation)ProtoNet · miniImageNet-modNovelGen vs BaseGen accuracyMixed
Use higher learning rate for FOMAML (ablation)FOMAML · FC100-modBaseGen accuracyImproved
Use lower learning rate for FOMAML (ablation)FOMAML · FC100-modNovelGen accuracyImproved
Fix support examples per base class (FixML) (ablation)SVM, Ridge, PN · CIFAR-FS-modNovelGen accuracyImproved
Fix support examples per base class (FixML) (ablation)SVM, Ridge, PN · CIFAR-FS-modBaseGen accuracyHurt
Reduce Zappos-ID training tasks from 1000 to 50 (ablation)PN vs SVM · Zappos-ID 2w5saccuracy gap PN over SVM: 86.58 -> 77.67Improved
Select snapshot by best ValGen instead of best NovelGenSVM, Ridge, PN, FOMAML · FC100NovelGen accuracy: 1.1, 1.2, 0.7, 0.9 lostHurt
Select snapshot by BaseGen instead of ValGenSVM · CIFAR-FS-modNovelGen accuracyHurt
Select snapshot by BaseGen instead of ValGenPN · tieredImageNet-modNovelGen accuracyImproved
Use ID validation (Zappos-ID, supervised IC) instead of OOD validationsnapshot selection · FSL benchmarksKendall rhoImproved
Use 160 novel classes instead of 20 for evaluationmethod comparison · miniImageNet/tieredImageNetconclusion flip and exaggeration frequencyImproved
Train on 351 base classes instead of 64 (tieredImageNet)PN vs Ridge · tieredImageNet-modconclusion flip rateImproved

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