Research map / Label-efficient and robust learning / Noisy labels
Beyond Images: Label Noise Transition Matrix Estimation for Tasks with Lower-Quality Features
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 |
|---|---|---|---|
| down-weight feature dimensions using f-mutual information and eigen-space decorrelation before HOC | HOC transition-matrix estimation · AG's news, DBpedia, Yelp-5, and Jigsaw (BERT) | estimation error | Helped |
| down-weight dimensions directly in the original feature space, without eigen-space projection (ablation) | HOC transition-matrix estimation · natural language benchmarks | estimation error | No clear effect |
| add eigen-space projection to decorrelate features before information-theoretic reweighting (ablation) | HOC transition-matrix estimation · tabular benchmarks | estimation error | No clear effect |
| add eigen-space projection before information-theoretic reweighting (ablation) | HOC transition-matrix estimation · natural language benchmarks | estimation error | Helped |
| use the estimated transition matrix in place of HOC's estimate for forward loss correction | downstream text classification · AG's news and DBpedia | clean test accuracy: 82.17 -> 85.01; 91.06 -> 97.77 | Helped |
| use TV-based f-mutual information instead of KL-based f-mutual information (ablation) | Ours-A · text and tabular benchmarks | estimation error | Mixed |
| use KL-based f-mutual information instead of TV-based f-mutual information (ablation) | Ours-X · text and tabular benchmarks | estimation error | Mixed |
Closest papers by meaning
- Estimating Noise Transition Matrix with Label Correlations for Noisy Multi-Label Learning (NeurIPS 2022 · similarity 0.90)
- Identifiability of Label Noise Transition Matrix (ICML 2023 · similarity 0.90)
- Learning the Latent Causal Structure for Modeling Label Noise (NeurIPS 2024 · similarity 0.89)
- Estimation of the Label-Noise Transition Matrix with Performance Guarantees via Selective Classification (NeurIPS 2026 · similarity 0.89)
- Learning Causal Transition Matrix for Instance-dependent Label Noise (AAAI 2025 · similarity 0.86)
- Estimating Noisy Class Posterior with Part-level Labels for Noisy Label Learning (CVPR 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.