Research map / Label-efficient and robust learning / Noisy labels
MIND: Decoupling Model-Induced Label Noise via Latent Manifold Disentanglement
Dayong Ren
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 global estimation (K=1) instead of multi-subspace estimation (ablation) | MIND · S3DIS (PCAM noise) | mIoU: 34.08 -> 40.58 | Improved |
| add orthogonality loss L_dec to gating (ablation) | MIND LDE · S3DIS (PCAM noise) | mIoU: 37.38 -> 40.58 | Improved |
| add momentum update to basis estimation (ablation) | MIND · S3DIS (PCAM noise) | mIoU: 38.12 -> 40.58 | Improved |
| use MIND instead of CA2C | noisy-label training · S3DIS PCAM | mIoU: 39.36 -> 40.58 | Improved |
| use MIND instead of DMLP meta-learning | noisy-label training · S3DIS PCAM | mIoU: 37.43 -> 40.58 | Improved |
| use GCE robust loss instead of standard CE | noisy-label training · ScanNetV2 PCAM | mIoU: 36.40 -> 17.99 | Hurt |
| use MIND instead of Standard CE | noisy-label training · S3DIS KPConv | mIoU: 63.51 -> 66.48 | Improved |
| use MIND for CIFAR transition matrix estimation instead of BLTM | IDN estimation · CIFAR-100 (40% IDN) | matrix estimation error: 0.35 -> 0.18 | Improved |
| use MIND instead of BLTM | CIFAR-100 training · CIFAR-100 (40% IDN) | accuracy: 43.50 -> 46.22 | Improved |
| apply MIND to OpenSeg pseudo-labels | OpenSeg · S3DIS | mIoU: 43.20 -> 51.34 | Improved |
| retrain OpenSeg pseudo-labels with standard CE | OpenSeg · S3DIS | mIoU: 43.20 -> 44.15 | No clear effect |
| use MIND instead of CA2C in zero-shot S3DIS to ScanNet transfer | noisy-label training · S3DIS->ScanNet PCAM | mIoU: 21.50 -> 23.54 | Improved |
Closest papers by meaning
- Instance-Dependent Label-Noise Learning With Manifold-Regularized Transition Matrix Estimation (CVPR 2022 · similarity 0.89)
- Learning the Latent Causal Structure for Modeling Label Noise (NeurIPS 2024 · similarity 0.87)
- NoiseGPT: Label Noise Detection and Rectification through Probability Curvature (NeurIPS 2024 · similarity 0.86)
- Learning Causal Transition Matrix for Instance-dependent Label Noise (AAAI 2025 · similarity 0.86)
- Learning with Open-world Noisy Data via Class-independent Margin in Dual Representation Space (AAAI 2025 · similarity 0.85)
- Learning from Noisy Labels via Conditional Distributionally Robust Optimization (NeurIPS 2024 · similarity 0.85)
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.
