AI Paper Atlas

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

Sinkhorn Label Allocation: Semi-Supervised Classification via Annealed Self-Training

Kai Sheng Tai, Peter D Bailis, Gregory Valiant

ICML 2021 · Label-efficient and robust learning › Semi-supervised learning · approach: Holistic semi-supervised pipelines · 7 citations (OpenAlex)

Read the paper

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
Use SLA label allocation instead of FixMatch pseudo-label thresholdingself-training · CIFAR-10, 40 labelstest error (%): 9.90±3.00 -> 5.17±0.32Improved
Use SLA label allocation instead of FixMatch pseudo-label thresholdingself-training · CIFAR-100, 2500 labelstest error (%): 27.99±0.42 -> 28.73±0.44Hurt
Use SLA label allocation instead of FixMatch pseudo-label thresholdingself-training · SVHN, 20 labelstest error (%): 14.92±7.82 -> 22.85±9.84Hurt
Use SLA label allocation instead of FixMatch pseudo-label thresholdingself-training · SVHN, 40 labelstest error (%): 4.74±3.28 -> 3.63±2.91Improved
Add class proportion upper bounds (ablation)SLA · CIFAR-10, 40 labels, Uniformtest error (%): 9.71±5.95 -> 5.17±0.32Improved
Add class proportion upper bounds estimated from labeled proportions (ablation)SLA · CIFAR-10, 40 labels, Multinomialtest error (%): 13.40±6.41 -> 14.95±7.12Hurt
Add a linear label-allocation ramp rather than allocating all examples from the start (ablation)SLA · CIFAR-10, 40 labelstest error (%): 13.67 ± 1.83 -> 5.17±0.32Improved
Increase Sinkhorn regularization from γ = 1 to γ = 10 (ablation)SLA · CIFAR-10, 40 labelstest error (%): 42.48 -> 5.78Improved
Increase Sinkhorn regularization from γ = 10 to γ = 100 (ablation)SLA · CIFAR-10, 40 labelstest error (%): 5.78 -> 4.94Improved
Increase Sinkhorn regularization from γ = 100 to γ = 1000 (ablation)SLA · CIFAR-10, 40 labelstest error (%): 4.94 -> 5.10Hurt
Use a truncated allocation schedule that stops before full allocationSLA · CIFAR-100test error (%)No clear effect

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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