Research map / Label-efficient and robust learning / Recognition and long-tail
Complementary-Label Learning for Arbitrary Losses and Models
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 Gradient Ascent (Free) instead of Forward (Fwd) | Complementary-label learning · MNIST | accuracy: 88.1\pm2.5\% vs 93.3\pm0.2\% | Hurt |
| Use Gradient Ascent (Free) instead of Forward (Fwd) | Complementary-label learning · Fashion-MNIST | accuracy: 78.7\pm1.4\% vs 82.7\pm0.4\% | Hurt |
| Use Gradient Ascent (Free) instead of Forward (Fwd) | Complementary-label learning · Kuzushiji-MNIST | accuracy: 63.8\pm1.1\% vs 64.1\pm0.4\% | No clear effect |
| Use Gradient Ascent (Free) instead of Forward (Fwd) | Complementary-label learning · CIFAR-10 | accuracy: 36.8\pm0.6\% vs 33.2\pm0.9\% | Helped |
| Use Free as the validation criterion instead of each method's own criterion | PC · MNIST | accuracy: 79.3\pm3.3\% vs 80.2\pm2.9\% | No clear effect |
| Use Free as the validation criterion instead of each method's own criterion | Forward (Fwd) · MNIST | accuracy: 93.3\pm0.2\% vs 92.2\pm0.6\% | No clear effect |
| Use Free as the validation criterion instead of each method's own criterion | PC · Fashion-MNIST | accuracy: 74.7\pm1.6\% vs 75.7\pm1.2\% | No clear effect |
| Use Free as the validation criterion instead of each method's own criterion | Forward (Fwd) · Fashion-MNIST | accuracy: 82.7\pm0.4\% vs 82.5\pm0.6\% | No clear effect |
| Use Free as the validation criterion instead of each method's own criterion | PC · Kuzushiji-MNIST | accuracy: 56.7\pm4.9\% vs 56.1\pm4.2\% | No clear effect |
| Use Free as the validation criterion instead of each method's own criterion | Forward (Fwd) · Kuzushiji-MNIST | accuracy: 64.1\pm0.4\% vs 63.9\pm1.9\% | No clear effect |
| Use Free as the validation criterion instead of each method's own criterion | PC · CIFAR-10 | accuracy: 33.4\pm2.0\% vs 25.9\pm7.6\% | Hurt |
| Use Free as the validation criterion instead of each method's own criterion | Forward (Fwd) · CIFAR-10 | accuracy: 33.2\pm0.9\% vs 33.7\pm1.1\% | No clear effect |
Closest papers by meaning
- Learning with Complementary Labels Revisited: The Selected-Completely-at-Random Setting Is More Practical (ICML 2024 · similarity 0.93)
- Forming Auxiliary High-confident Instance-level Loss to Promote Learning from Label Proportions (CVPR 2025 · similarity 0.86)
- Label Noise: Ignorance Is Bliss (NeurIPS 2024 · similarity 0.86)
- Gradient Aligned Regression via Pairwise Losses (ICML 2025 · 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.