AI Paper Atlas

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

No One Representation to Rule Them All: Overlapping Features of Training Methods

Raphael Gontijo-Lopes, Yann Dauphin, Ekin Dogus Cubuk

ICLR 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Data augmentation and generalization bias · 5 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
Combine models trained with increasingly different methodologiesResNet-50 ensembles · ImageNetensemble accuracy: 76.5% -> 83.4%Improved
Add a lower-accuracy model trained with a different objective and datasetResNet-50 · ImageNettop-1 accuracy: 76.2% -> 77.7%Improved
Greedily add only lower-accuracy modelsBiT-1k ensemble · ImageNetensemble accuracy: 82.85% -> 86.66%Improved
Train models with more divergent methodologies rather than only changing initializationModel pairs relative to ResNet-50 · ImageNeterror inconsistency: 0.0850 -> 0.2295Improved
Combine embeddings from differently trained models rather than use either embedding aloneImageNet linear evaluation · ImageNetLBFGS accuracyImproved
Use more divergent training methodologies to produce representations, then select diverse dimensionsImageNet feature representations · ImageNetfeatures needed for 76% accuracy: 2304 -> 512Improved
Combine features from differently trained modelsStacked generalization · Pascal VOC11-point mAP: 91.4%Improved
Use class-aware interpolation for ResNet-50 and CLIP-S logitsResNet-50 + CLIP-S ensemble · ImageNetaccuracy: 80.424% -> 80.452%Improved

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