AI Paper Atlas

Research map / Label-efficient and robust learning / Semi-supervised learning

ExMatch: Self-guided Exploitation for Semi-Supervised Learning with Scarce Labeled Samples

Noo-ri Kim, Jin-Seop Lee, Jee-Hyong Lee

ECCV 2024 · Label-efficient and robust learning › Semi-supervised learning · approach: Confidence-based pseudo-label selection · 2 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 self-guided exploitation loss selected by confidence and self-pseudo-label consistency, with weak augmentation (ablation)FixMatch · CIFAR-10 (10 labels, 3 folds)accuracy: 60.84 -> 74.83Improved
only add self-pseudo-label consistency to the FixMatch unsupervised loss (ablation)FixMatch · CIFAR-10 (10 labels, 3 folds)accuracy: 60.84 -> 64.86No clear effect
exploit all pseudo-labeled samples without selection (ablation)ExMatch · CIFAR-10 (10 labels, 3 folds)accuracy: 74.83 -> 12.31Hurt
select exploited samples by confidence only (ablation)ExMatch · CIFAR-10 (10 labels, 3 folds)accuracy: 74.83 -> 29.25Hurt
add confidence threshold on top of self-pseudo-label consistency (ablation)ExMatch · CIFAR-10 (10 labels, 3 folds)accuracy: 56.61 -> 74.83Improved
use weak instead of strong augmentation in the exploitation loss (ablation)ExMatch · CIFAR-10 (10 labels, 3 folds)accuracy: 50.07 -> 74.83Improved
use similarity-based self-pseudo-labels instead of fine-tuning the self-supervised model (ablation)ExMatch · CIFAR-10 (10 labels)accuracy: 67.82 -> 74.83Improved
use a weaker self-supervised model (RotNet NiN-4) (ablation)ExMatch · CIFAR-10 (10 labels)accuracy: 74.83 -> 57.91Hurt
use smaller self-supervised network or a model trained on ImageNet-32 (ablation)ExMatch · CIFAR-10 (10 labels)accuracy: 74.83 vs 74.72 (ResNet-18) and 74.85 (ImageNet-32)No clear effect
replace FixMatch with ExMatch on CIFAR-100 with 100 labelsExMatch · CIFAR-100 (100 labels)accuracy: 24.11 -> 32.49Improved
replace FlexMatch/FreeMatch with ExMatch on CIFAR-10 with 20 and 40 labelsExMatch · CIFAR-10 (20 and 40 labels)accuracy: 94.62 vs 84.66 (20 labels); 94.88 vs 89.78 (40 labels)Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 4 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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