Research map / Label-efficient and robust learning / Noisy labels
Robust Inference via Generative Classifiers for Handling Noisy Labels
Kimin Lee, Sukmin Yun, Kibok Lee, Honglak Lee, Bo Li, Jinwoo Shin
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 |
|---|---|---|---|
| Replace softmax inference with ensemble RoG | ResNet · CIFAR-10, uniform noise | test accuracy: 80.88 / 80.88 -> 84.62; 61.98 -> 75.28; 39.96 -> 54.57 | Improved |
| Replace softmax inference with ensemble RoG | ResNet · CIFAR-100, uniform noise | test accuracy: 64.43 / 68.29; 48.62 / 60.76; 27.57 / 48.42 | Improved |
| Replace softmax inference with ensemble RoG | ResNet · SVHN, uniform noise | test accuracy: 83.52 / 91.67; 72.89 / 87.16; 61.23 / 80.52 | Improved |
| Add MCD parameter estimation instead of naive sample estimation (ablation) | Generative classifier · CIFAR-10, Uniform (40%), DenseNet | test accuracy: 76.75 -> 80.27 | Improved |
| Add layer-wise ensemble to sample-estimated generative classifiers (ablation) | Generative classifier · CIFAR-10, Uniform (40%), DenseNet | test accuracy: 76.75 -> 81.27 | Improved |
| Add layer-wise ensemble to MCD-estimated generative classifiers (ablation) | RoG · CIFAR-10, Uniform (40%), DenseNet | test accuracy: 80.27 -> 81.83 | Improved |
| Add MCD estimation instead of sample estimation (ablation) | Ensemble generative classifier · CIFAR-10, Uniform (40%), DenseNet | test accuracy: 81.27 -> 81.83 | Improved |
| Add RoG inference to Co-teaching | Co-teaching · CIFAR-100, Flip (45%) | test accuracy: 33.34 -> 43.18 | Improved |
| Replace softmax inference with RoG | DenseNet · CIFAR-100, open-set noise | test accuracy: 79.01 -> 83.37 | Improved |
| Use LDA tied covariance rather than identity covariance (ablation) | Generative classifier · CIFAR-10, uniform noise, ResNet-34 | test accuracy | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Co-teaching — Co-teaching: Robust training of deep neural networks with extremely noisy labels (NeurIPS 2018) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- API-Net: Robust Generative Classifier via a Single Discriminator (ECCV 2020 · similarity 0.88)
- Label-Noise Robust Generative Adversarial Networks (CVPR 2019 · similarity 0.87)
- Understanding the Limitations of Conditional Generative Models (ICLR 2020 · similarity 0.87)
- Generative Classifiers Avoid Shortcut Solutions (ICLR 2025 · similarity 0.87)
- Fusing Conditional Submodular GAN and Programmatic Weak Supervision (AAAI 2024 · similarity 0.86)
- Training Deep Neural Networks with Virtual Smoothing Classes (AAAI 2025 · 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.
