Research map / Label-efficient and robust learning / Recognition and long-tail
On Statistical Bias In Active Learning: How and When to Fix It
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 |
|---|---|---|---|
| Use LURE or PURE corrective weighting instead of the unweighted active-learning estimator | Active-learning risk estimation · Pool-based active learning | risk-estimation bias | Helped |
| Use LURE or PURE to estimate risk with a BALD-style acquisition proposal | Active-learning risk estimation · Modified MNIST | estimator variance | Helped |
| Use LURE or PURE instead of the unweighted estimator when training linear regression | Linear regression · Toy nonlinear data | test loss | Helped |
| Use LURE or PURE instead of the unweighted estimator when training an overparameterized BNN | Convolutional BNN · Modified MNIST | test loss | Hurt |
| Use LURE or PURE instead of the unweighted estimator when training an overparameterized BNN | Convolutional BNN · Modified MNIST | test accuracy | Hurt |
| Use LURE or PURE instead of the unweighted estimator when training a neural model | Neural models · FashionMNIST, MNIST with Monte Carlo dropout, and balanced MNIST | test NLL | Mixed |
| Use LURE rather than PURE | Unbiased active-learning risk estimation · Pool-based active learning | estimator variance | Helped |
| Use an acquisition proposal proportional to non-negative loss | LURE and PURE · Pool-based active learning | risk-estimation variance | Helped |
| Retain the bias of the unweighted estimator during neural-network training rather than correcting it | Overparameterized neural networks · Active-learning training | generalization performance | Helped |
Closest papers by meaning
- Debiased Active Learning with Variational Gradient Rectifier (AAAI 2025 · similarity 0.87)
- Active Statistical Inference (ICML 2024 · similarity 0.87)
- Uncertainty Herding: One Active Learning Method for All Label Budgets (ICLR 2025 · similarity 0.86)
- Improved Algorithm for Deep Active Learning under Imbalance via Optimal Separation (ICML 2025 · similarity 0.86)
- Towards Cost-Effective Learning: A Synergy of Semi-Supervised and Active Learning (CVPR 2025 · similarity 0.86)
- Transductive Active Learning: Theory and Applications (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.