Research map / Learning theory and algorithms / Prediction and decision losses
Towards Robust and Reproducible Active Learning Using Neural Networks
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 |
|---|---|---|---|
| Add strong regularization (random augmentation and SWA) to random sampling (ablation) | RSB · CIFAR10, 10% labeled | test accuracy: 69.16 -> 82.16 | Helped |
| Add strong regularization (random augmentation and SWA) to random sampling (ablation) | RSB · CIFAR100, 10% labeled | test accuracy: 34.73 -> 39.44 | Helped |
| Add random augmentation (ablation) | RSB · CIFAR10, 10% labeled | test accuracy: 69.16 -> 76.36 | Helped |
| Add SWA (ablation) | RSB · CIFAR10, 10% labeled | test accuracy: 69.16 -> 74.6 | Helped |
| Add Shake-Shake regularization (ablation) | RSB · CIFAR100, 10% labeled | test accuracy: 34.73 -> 39.23 | Helped |
| Use active learning rather than random sampling | VGG16 classifier · CIFAR10 and CIFAR100 across labeled-data fractions | test accuracy | Mixed |
| Increase labeled-data fraction from 20% to 30% | UC · CIFAR10 | test accuracy: 0.54% | Hurt |
| Use a strongly regularized model instead of the standard model (ablation) | RSB · CIFAR10, 40% labeled | test accuracy: 80.88 -> 89.13 | Helped |
| Use a strongly regularized model instead of the standard model (ablation) | RSB · CIFAR10 with 20% oracle noise, 30% labeled | test accuracy: 75.89 -> 85.9 | Helped |
| Change the classifier architecture used to evaluate selected active sets | CIFAR10 active-learning methods · CIFAR10 transferability | test accuracy | Mixed |
Closest papers by meaning
- Provably Neural Active Learning Succeeds via Prioritizing Perplexing Samples (ICML 2024 · similarity 0.89)
- Neural Active Learning Beyond Bandits (ICLR 2024 · similarity 0.89)
- Transductive Active Learning: Theory and Applications (NeurIPS 2024 · similarity 0.88)
- Active Learning Guided by Efficient Surrogate Learners (AAAI 2024 · similarity 0.88)
- Robust Sampling for Active Statistical Inference (NeurIPS 2025 · similarity 0.87)
- Instance-wise Supervision-level Optimization in Active Learning (CVPR 2025 · similarity 0.87)
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.