AI Paper Atlas

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

Are labels informative in semi-supervised learning? Estimating and leveraging the missing-data mechanism.

Aude Sportisse, Hugo Schmutz, Olivier HUMBERT, Charles Bouveyron, Pierre-Alexandre Mattei

ICML 2023 · Label-efficient and robust learning › Semi-supervised learning · approach: Holistic semi-supervised pipelines · 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
Use class-specific missingness estimates and IPW instead of the original pseudo-label methodMNIST semi-supervised learning · MNIST (unbalanced labels)test accuracy: 92.95 -> 94.83Improved
Use class-specific missingness estimates and IPW instead of the original pseudo-label methodMNIST semi-supervised learning · MNIST (unbalanced labels)test loss: 0.141 -> 0.103Improved
Estimate class-specific label missingness and debias FixMatch with IPWFixMatch · CIFAR-10 (informative labels)test accuracy: 90.91 -> 91.88Improved
Estimate class-specific label missingness and debias FixMatch with IPWFixMatch · CIFAR-10 (informative labels)test loss: 0.426 -> 0.321Improved
Estimate the missing-data mechanism and debias with IPW instead of using pseudo-labeling aloneSemi-supervised learning · dermaMNIST (informative labels)test accuracy: 57.72 -> 66.65Improved
Estimate the missing-data mechanism by maximum likelihood and use IPWSemi-supervised learning · noduleMNIST3D (informative labels)test accuracy: 84.91 -> 85.8Improved
Estimate class-specific label missingness rather than relying on class proportionsSemi-supervised learning · dermaMNIST (informative labels)Accuracy Naevi: 66.14\pm 5.86 -> 93.54\pm2.30Improved
Use the moment estimator with gradients propagated through class proportions instead of the maximum-likelihood estimatorSemi-supervised learning · noduleMNIST3D (informative labels)test accuracy: 85.8 -> 82.26Hurt

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