Research map / Label-efficient and robust learning / Semi-supervised learning
WCP: Worst-Case Perturbations for Semi-Supervised Deep Learning
Check 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 |
|---|---|---|---|
| Use worst-case additive and DropConnect perturbations | 13-layer semi-supervised classifier · CIFAR-10, 1,000–4,000 labels | error rate: 46.43±1.21 -> 17.62±1.52; 33.94±0.73 -> 11.93±0.39; 20.66±0.57 -> 9.72±0.31 | Helped |
| Use worst-case additive and DropConnect perturbations | 13-layer semi-supervised classifier · SVHN, 250–1,000 labels | error rate: 27.77±3.18 -> 4.29±0.10; 16.88±1.30 -> 3.75±0.11; 12.32±0.95 -> 3.58±0.186 | Helped |
| Add additive perturbation (ablation) | WCP on CIFAR-10 · CIFAR-10, 4,000 labels | error rate: 10.15 -> 9.51 | Helped |
| Add DropConnect perturbation (ablation) | WCP on CIFAR-10 · CIFAR-10, 4,000 labels | error rate: 9.85 -> 9.51 | Helped |
| Add entropy minimization (ablation) | WCP on CIFAR-10 · CIFAR-10, 4,000 labels | error rate: 9.85 -> 9.51 | Helped |
| Move DropConnect from the second to the first layer of each block (ablation) | WCP on CIFAR-10 · CIFAR-10, 4,000 labels | error rate: 9.51 -> 9.77 | Hurt |
| Move DropConnect from the second to the third layer of each block (ablation) | WCP on CIFAR-10 · CIFAR-10, 4,000 labels | error rate: 9.51 -> 10.08 | Hurt |
| Set DropConnect ratio to 0.1 rather than 0.2 (ablation) | WCP on CIFAR-10 · CIFAR-10, 4,000 labels | error rate: 9.51 -> 9.81 | Hurt |
| Set DropConnect ratio to 0.3 rather than 0.2 (ablation) | WCP on CIFAR-10 · CIFAR-10, 4,000 labels | error rate: 9.51 -> 9.66 | Hurt |
| Set DropConnect ratio to 0.4 rather than 0.2 (ablation) | WCP on CIFAR-10 · CIFAR-10, 4,000 labels | error rate: 9.51 -> 9.78 | Hurt |
| Set DropConnect ratio to 0.5 rather than 0.2 (ablation) | WCP on CIFAR-10 · CIFAR-10, 4,000 labels | error rate: 9.51 -> 9.92 | Hurt |
| Set DropConnect ratio to 0.7 rather than 0.2 (ablation) | WCP on CIFAR-10 · CIFAR-10, 4,000 labels | error rate: 9.51 -> 10.26 | Hurt |
Closest papers by meaning
- Regularization With Stochastic Transformations and Perturbations for Deep Semi-Supervised Learning (NeurIPS 2016 · similarity 0.87)
- Adversarial Weight Perturbation Improves Generalization in Graph Neural Networks (AAAI 2023 · similarity 0.87)
- Semi-Supervised Learning with Variational Bayesian Inference and Maximum Uncertainty Regularization (AAAI 2021 · similarity 0.87)
- On the Inherent Regularization Effects of Noise Injection During Training (ICML 2021 · similarity 0.86)
- Generalizable Adversarial Training via Spectral Normalization (ICLR 2019 · similarity 0.86)
- Normalize Then Propagate: Efficient Homophilous Regularization for Few-Shot Semi-Supervised Node Classification (AAAI 2025 · 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.