Research map / Learning theory and algorithms / Prediction and decision losses
Bounding Uncertainty for Active Batch Selection
Hanmo Wang, Runwu Zhou, Yi-Dong Shen
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 LBC with random greedy solver instead of QP-based MMD | batch mode active learning · RCV1 | running time: 7458.11 s -> 9.66 s (772.08x) | Improved |
| use LBC instead of MCDR, RPE, Batchrank | batch mode active learning · 15 datasets | running time: e.g. TDT2 5251.77 s (MCDR) -> 195.31 s | Improved |
| use LBC instead of random selection | batch mode active learning · 15 datasets | accuracy win rate: e.g. 99/0 on RCV1, 96/0 on segmentation | Improved |
| use lower-bounded certainty instead of direct approach (no bound) (ablation) | LBC · RCV1, segmentation, 20News | accuracy win/loss: 91/0 on RCV1; 92/0 on 20News | Improved |
| use lower-bounded certainty instead of direct approach (no bound) (ablation) | LBC · waveform, twonorm | accuracy win/loss: 0/4 on waveform; 0/1 on twonorm | Hurt |
| use lower-bounded certainty instead of screening approach (ablation) | LBC · waveform, twonorm, 20News | accuracy win/loss: 90/0 waveform; 97/0 twonorm; 98/0 20News | Improved |
| use LBC instead of direct or screening variants (ablation) | LBC · Reuters, ORL | accuracy: 0/0 and 0/0 vs direct on ORL; 0/0 vs direct on Reuters | No clear effect |
Closest papers by meaning
- Batch Active Learning at Scale (NeurIPS 2021 · similarity 0.89)
- Uncertainty for Active Learning on Graphs (ICML 2024 · similarity 0.86)
- Leveraging Uncertainty Estimates To Improve Classifier Performance (ICLR 2024 · similarity 0.86)
- Uncertainty Herding: One Active Learning Method for All Label Budgets (ICLR 2025 · similarity 0.86)
- Provably Neural Active Learning Succeeds via Prioritizing Perplexing Samples (ICML 2024 · similarity 0.85)
- Querying Easily Flip-flopped Samples for Deep Active Learning (ICLR 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.
