Research map / Label-efficient and robust learning / Recognition and long-tail
CCMN: A General Framework for Learning With Class-Conditional Multi-Label Noise
Ming-Kun Xie, Sheng-Jun Huang
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 unbiased estimator (Hamming loss, Square) instead of baseline loss | Lh/Square, linear · yeast CCMN | Hamming loss: .280 -> .218 | Improved |
| use unbiased estimator (ranking, Sigmoid) instead of baseline loss | Lr/Sigmoid, linear · yeast CCMN | ranking loss: .261 -> .204 | Improved |
| use unbiased estimator (ranking, Sigmoid) instead of baseline loss | Lr/Sigmoid, linear · yeast CCMN | average precision: .651 -> .720 | Improved |
| use unbiased estimator (ranking, Sigmoid) instead of baseline loss | Lr/Sigmoid, linear · scene CCMN | average precision: .758 -> .809 | Improved |
| use unbiased estimator (ranking, Sigmoid) instead of baseline loss | Lr/Sigmoid, linear · mirflickr CCMN | average precision: .823 -> .852 | Improved |
| use unbiased estimator (ranking, Sigmoid) instead of baseline loss | Lr/Sigmoid, MLP · scene CCMN | average precision: .752 -> .816 | Improved |
| use unbiased estimator (Hamming, Hinge) instead of baseline loss | Lh/Hinge, MLP · mirflickr CCMN | Hamming loss: .124 -> .106 | Improved |
| use unbiased estimator uPML instead of PMLNI | PML data · yeast | Hamming loss: .269 -> .210 | Improved |
| use unbiased estimator uPML instead of fPML | PML data · yeast | Hamming loss: .303 -> .208 | Improved |
| use uPML Lr instead of PMLNI | PML data · enron | average precision: .270 -> .563 | Improved |
| use uPML instead of PMLNI | PML data · music_style | average precision: .723 -> .716 | Hurt |
Closest papers by meaning
- Learning from Noisy Labels with No Change to the Training Process (ICML 2021 · similarity 0.92)
- Noise-Robust Learning from Multiple Unsupervised Sources of Inferred Labels (AAAI 2022 · similarity 0.91)
- Beyond Class-Conditional Assumption: A Primary Attempt to Combat Instance-Dependent Label Noise (AAAI 2021 · similarity 0.90)
- Limited-Supervised Multi-Label Learning with Dependency Noise (AAAI 2024 · similarity 0.89)
- Addressing Multi-Label Learning with Partial Labels: From Sample Selection to Label Selection (AAAI 2025 · similarity 0.88)
- Label Noise: Ignorance Is Bliss (NeurIPS 2024 · similarity 0.87)
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.
