Research map / Label-efficient and robust learning / Noisy labels
Dual T: Reducing Estimation Error for Transition Matrix in Label-noise Learning
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 |
|---|---|---|---|
| Replace T with dual-T in transition-matrix-dependent label-noise algorithms | Coteaching, MentorNet, Forward, Reweighting, and Revision · MNIST, F-MNIST, CIFAR10, and CIFAR100 across Sym-20%, Sym-50%, and Pair-45% | accuracy: MNIST Coteaching Sym-20%: 93.41 -> 97.52; Sym-50%: 84.13 -> 83.20; Pair-45%: 63.60 -> 86.78 | Mixed |
| Replace T with dual-T in Coteaching | Coteaching · CIFAR10, Sym-20% | accuracy: 79.47 ±0.20 -> 90.37* ±0.12 | Helped |
| Replace T with dual-T in Forward | Forward · CIFAR100, Pair-45% | accuracy: 25.10 ±0.77 -> 26.08 ±0.93 | Helped |
| Replace T with dual-T in Reweighting | Reweighting · CIFAR100, Pair-45% | accuracy: 25.86 ±0.55 -> 25.84±0.42 | Hurt |
| Replace T with dual-T in Revision | Revision · Clothing1M | accuracy: 71.01 -> 71.49* | Helped |
| Estimate the transition matrix with dual-T rather than T | Transition-matrix estimator · CIFAR100, small training sample | transition-matrix estimation error | Hurt |
| Factor the transition matrix through an intermediate class instead of estimating it directly | Transition-matrix estimator · Synthetic data and real-world image datasets | transition-matrix estimation error | Helped |
Closest papers by meaning
- Estimation of the Label-Noise Transition Matrix with Performance Guarantees via Selective Classification (NeurIPS 2026 · similarity 0.93)
- Learning the Latent Causal Structure for Modeling Label Noise (NeurIPS 2024 · similarity 0.90)
- Estimating Noisy Class Posterior with Part-level Labels for Noisy Label Learning (CVPR 2024 · similarity 0.90)
- Learning Causal Transition Matrix for Instance-dependent Label Noise (AAAI 2025 · similarity 0.89)
- Dirichlet-based Per-Sample Weighting by Transition Matrix for Noisy Label Learning (ICLR 2024 · 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.