Research map / Label-efficient and robust learning / Semi-supervised learning
Regularization With Stochastic Transformations and Perturbations for Deep Semi-Supervised Learning
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 transformation/stability loss to labeled-only training (ablation) | sparse convnet · MNIST (100 labels) | error rate (%): 5.44 -> 0.76 | Helped |
| add mutual-exclusivity loss to labeled-only training (ablation) | sparse convnet · MNIST (100 labels) | error rate (%): 5.44 -> 3.92 | Helped |
| add both unsupervised losses (ablation) | sparse convnet · MNIST (100 labels) | error rate (%): 5.44 -> 0.55 | Helped |
| add both unsupervised losses (ablation) | sparse convnet · MNIST (all labels) | error rate (%): 0.32 -> 0.27 | Helped |
| add both unsupervised losses | sparse convnet · SVHN (1% labels) | error rate (%): 12.25 -> 6.03 | Helped |
| add both unsupervised losses | sparse convnet · SVHN (100% labels) | error rate (%): 2.28 -> 2.22 | Helped |
| add both unsupervised losses | sparse convnet · NORB (1% labels) | error rate (%): 10.01 -> 2.15 | Helped |
| add both unsupervised losses | sparse convnet · NORB (100% labels) | error rate (%): 1.63 -> 1.63 | No clear effect |
| add both unsupervised losses | sparse convnet · CIFAR10 (4000 labels) | error rate (%): 13.60 -> 11.29 | Helped |
| add both unsupervised losses | AlexNet (cuda-convnet) · ILSVRC 2012 (10% labels) | error rate (%): 45.91 -> 39.84 | Helped |
| add unsupervised losses with 1% SVHN labels using cuda-convnet, transformation/stability + mutual-exclusivity | cuda-convnet · SVHN (1-100% labels) | error rate | Helped |
| add transformation/stability + mutual-exclusivity on all CIFAR100 labels with no augmentation | sparse convnet · CIFAR100 (all labels) | error rate (%): 23.82 -> 21.43 | Helped |
Closest papers by meaning
- Big Self-Supervised Models are Strong Semi-Supervised Learners (NeurIPS 2020 · similarity 0.89)
- Semi-supervised Deep Kernel Learning: Regression with Unlabeled Data by Minimizing Predictive Variance (NeurIPS 2018 · similarity 0.88)
- WCP: Worst-Case Perturbations for Semi-Supervised Deep Learning (CVPR 2020 · similarity 0.87)
- Matrix Completion for Graph-Based Deep Semi-Supervised Learning (AAAI 2019 · similarity 0.87)
- CGMatch: A Different Perspective of Semi-supervised Learning (CVPR 2025 · similarity 0.86)
- Data Augmentation with Diffusion for Open-Set Semi-Supervised Learning (NeurIPS 2024 · 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.