Research map / Label-efficient and robust learning / Recognition and long-tail
Learning from Complementary Labels
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 |
|---|---|---|---|
| Train from complementary labels with PC and sigmoid loss instead of the partial-label baseline | Classifier · SATIMAGE (1–7) | accuracy (%): 60.7(3.7) -> 68.7(5.4) | Helped |
| Train from complementary labels with PC and sigmoid loss instead of the multi-label baseline | Classifier · PENDIGITS (1–10) | accuracy (%): 16.1(4.6) -> 38.0(4.3) | Helped |
| Use PC/sigmoid instead of partial-label or multi-label learning | Classifier · WAVEFORM1 (1–3) | accuracy (%): 85.7(0.9) -> 85.8(0.5) | No clear effect |
| Use PC/sigmoid instead of partial-label or multi-label learning | Classifier · LETTER (1–25) | accuracy (%): 5.0(1.0) -> 5.1(2.1) | No clear effect |
| Combine ordinary and complementary labeled data (α=1/2) rather than use ordinary labels alone | PC/sigmoid classifier · LETTER (1–25) | accuracy (%): 11.9(1.7) -> 31.0(1.7) | Helped |
| Combine ordinary and complementary labeled data (α=1/2) rather than use complementary labels alone | PC/sigmoid classifier · USPS (1–10) | accuracy (%): 28.5(3.6) -> 59.3(2.2) | Helped |
| Replace sigmoid with ramp loss in OVA (ablation) | MNIST complementary-label classifier · MNIST (8 classes) | accuracy (%): 73.9 (1.2) -> 71.4 (4.0) | No clear effect |
| Replace OVA with PC while retaining sigmoid loss (ablation) | MNIST complementary-label classifier · MNIST (8 classes) | accuracy (%): 73.9 (1.2) -> 72.9 (3.0) | No clear effect |
| Use PC with ramp instead of OVA with sigmoid (ablation) | MNIST complementary-label classifier · MNIST (8 classes) | accuracy (%): 73.9 (1.2) -> 71.4 (2.4) | No clear effect |
Closest papers by meaning
- Complementary-Label Learning for Arbitrary Losses and Models (ICML 2019 · similarity 0.94)
- Learning with Complementary Labels Revisited: The Selected-Completely-at-Random Setting Is More Practical (ICML 2024 · similarity 0.93)
- Learning with Multiple Complementary Labels (ICML 2020 · similarity 0.92)
- Learning with Partial-Label and Unlabeled Data: A Uniform Treatment for Supervision Redundancy and Insufficiency (ICML 2024 · similarity 0.86)
- Imprecise Label Learning: A Unified Framework for Learning with Various Imprecise Label Configurations (NeurIPS 2024 · similarity 0.86)
- Semi-supervised Multi-label Learning with Balanced Binary Angular Margin Loss (NeurIPS 2024 · similarity 0.86)
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.