Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Semi-Supervised Learning with Variational Bayesian Inference and Maximum Uncertainty Regularization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Add variational dropout to Mean Teacher | Mean Teacher · CIFAR-10, 1000 labels | classification error: 19.63$\pm$0.33 -> 16.35$\pm$0.18 | Helped |
| Add maximum uncertainty regularization to Mean Teacher | Mean Teacher · CIFAR-10, 1000 labels | classification error: 19.63$\pm$0.33 -> 17.96$\pm$0.32 | Helped |
| Add both variational dropout and MUR to Mean Teacher | Mean Teacher · CIFAR-10, 1000, 2000, and 4000 labels | classification error: 19.63$\pm$0.33 -> 15.47$\pm$0.13; 15.07$\pm$0.10 -> 10.57$\pm$0.28; 11.65$\pm$0.09 -> 8.54$\pm$0.20 | Helped |
| Add variational dropout to ICT | ICT · CIFAR-10, 1000, 2000, and 4000 labels | classification error: 14.15$\pm$0.16 -> 10.13$\pm$0.21; 11.56$\pm$0.07 -> 8.83$\pm$0.15; 9.18$\pm$0.03 -> 7.48$\pm$0.11 | Helped |
| Add MUR to ICT | ICT · CIFAR-10 and CIFAR-100 | classification error: 13.54$\pm$0.23; 10.49$\pm$0.07; 8.55$\pm$0.06; 34.91$\pm$0.20 | Helped |
| Add MUR on top of variational dropout | ICT · CIFAR-10 and CIFAR-100 | classification error: ICT + VD: 10.13$\pm$0.21, 8.83$\pm$0.15, 7.48$\pm$0.11, 34.12$\pm$0.16; ICT + VD + MUR: 10.37$\pm$0.25, 8.79$\pm$0.16, 7 | No clear effect |
| Add variational dropout to Mean Teacher | Mean Teacher · SVHN, 250, 500, and 1000 labels | classification error: 5.57$\pm$1.52 -> 5.26$\pm$1.73; 3.86$\pm$0.15 -> 3.39$\pm$0.10; 3.72$\pm$0.10 -> 3.28$\pm$0.08 | Helped |
| Add MUR to Mean Teacher | Mean Teacher · SVHN, 250, 500, and 1000 labels | classification error: 5.57$\pm$1.52 -> 6.45$\pm$1.29; 3.86$\pm$0.15 -> 3.66$\pm$0.07; 3.72$\pm$0.10 -> 3.48$\pm$0.04 | Mixed |
| Add both variational dropout and MUR to Mean Teacher | Mean Teacher · SVHN, 250, 500, and 1000 labels | classification error: 5.57$\pm$1.52 -> 6.82$\pm$2.01; 3.86$\pm$0.15 -> 3.21$\pm$0.13; 3.72$\pm$0.10 -> 3.16$\pm$0.07 | Mixed |
| Add MUR to Pi-model | Pi-model · SVHN, 250 labels | classification error: 13.37$\pm$0.97 -> 15.04$\pm$0.75 | Hurt |
| Add MUR to Pi-model | Pi-model · SVHN, 500 and 1000 labels | classification error: 7.25$\pm$0.36 -> 5.43$\pm$0.27; 5.18$\pm$0.13 -> 4.15$\pm$0.10 | Helped |
| Use random rather than maximum-uncertainty virtual points (ablation) | Mean Teacher + MUR · CIFAR-10, 1000 labels | classification error | Hurt |
Closest papers by meaning
- There Are Many Consistent Explanations of Unlabeled Data: Why You Should Average (ICLR 2019 · similarity 0.88)
- Towards Realistic Long-Tailed Semi-Supervised Learning: Consistency Is All You Need (CVPR 2023 · similarity 0.88)
- Variational Bayesian Last Layers (ICLR 2024 · similarity 0.86)
- SimPro: A Simple Probabilistic Framework Towards Realistic Long-Tailed Semi-Supervised Learning (ICML 2024 · similarity 0.86)
- Unknown-Aware Graph Regularization for Robust Semi-supervised Learning from Uncurated Data (AAAI 2024 · similarity 0.86)
- Maxitive Donsker-Varadhan Formulation for Possibilistic Variational Inference (NeurIPS 2026 · similarity 0.85)
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.