Research map / Learning theory and algorithms / Prediction and decision losses
Learning with Labeling Induced Abstentions
Kareem Amin, Giulia DeSalvo, Afshin Rostamizadeh
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 mixture requester (KNN + margin) instead of margin requester | DPL-Simplified · 10 binary datasets | online mistakes / conditional test error: MIX better on 8 of 10 datasets | Improved |
| MIX vs margin requester on cod-rna | DPL-Simplified · cod-rna | online mistakes: 360 ± 13 -> 329 ± 12 | Improved |
| MIX vs margin requester on cod-rna | DPL-Simplified · cod-rna | conditional test error (%): 2.2 ± 0.06 -> 1.5 ± 0.07 | Improved |
| MIX vs margin requester on ijcnn | DPL-Simplified · ijcnn | conditional test error (%): 2.4 ± 0.16 -> 0.7 ± 0.14 | Improved |
| MIX vs margin requester on skin | DPL-Simplified · skin | online mistakes: 662 ± 110 -> 36 ± 31 | Improved |
| MIX vs margin requester on HIGGS | DPL-Simplified · HIGGS | online mistakes: 5500 ± 90 -> 5689 ± 69 | Hurt |
| MIX vs margin requester on a9a | DPL-Simplified · a9a | conditional test error (%): 8.9 ± 0.44 -> 10.3 ± 0.30 | Hurt |
| MIX vs margin requester on cifar | DPL-Simplified · cifar | conditional test error (%): 5.1 ± 0.14 -> 3.8 ± 0.09 | Improved |
| MIX vs KNN requester on a9a | DPL-Simplified · a9a | online mistakes: 2017 ± 35 -> 1713 ± 51 | Improved |
| MIX vs KNN requester on HIGGS | DPL-Simplified · HIGGS | conditional test error (%): 35.2 ± 0.31 -> 34.0 ± 0.27 | Improved |
| MIX vs margin requester on shuttle | DPL-Simplified · shuttle | conditional test error (%): 0.8 ± 0.09 -> 0.1 ± 0.05 | Improved |
| replace non-trivial requester r_t with uniform random requester (ablation) | DPL-IWAL · six LIBSVM datasets, online mistakes and conditional loss | online mistakes / conditional loss | Hurt |
Closest papers by meaning
- Active Learning from Imperfect Labelers (NeurIPS 2016 · similarity 0.85)
- Active Sampling for Text Classification with Subinstance Level Queries (AAAI 2022 · similarity 0.85)
- A Label is Worth A Thousand Images in Dataset Distillation (NeurIPS 2024 · similarity 0.84)
- Imprecise Label Learning: A Unified Framework for Learning with Various Imprecise Label Configurations (NeurIPS 2024 · similarity 0.83)
- Uncertainty Herding: One Active Learning Method for All Label Budgets (ICLR 2025 · similarity 0.83)
- Divide and Conquer: Learning Label Distribution with Subtasks (ICML 2025 · similarity 0.83)
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.
