Research map / Continual learning / Class-incremental
Imbalanced Deep Learning by Minority Class Incremental Rectification
Qi Dong, Shaogang Gong, Xiatian Zhu
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 |
|---|---|---|---|
| add CRL to the base network | DeepID2 · CelebA | mean class-balanced accuracy (%): 81 -> 87 | Improved |
| add CRL to the base network | MTCT · X-Domain | mean class-balanced accuracy (%): 73.53 -> 80.42 | Improved |
| add CRL to ResNet50 | ResNet50 · DeepFashion | mean class-balanced accuracy (%): 52.20 -> 54.56 | Improved |
| add CRL to CNNs on balanced CIFAR-100 | CifarNet/ResNet32/DenseNet · CIFAR-100 | accuracy (%): +3.6, +1.2, +0.8 | Improved |
| add CRL under simulated imbalance | ResNet32 · CIFAR-100 imbalanced (gamma=1) | mean class-balanced accuracy (%): 42.1 -> 45.2 | Improved |
| add hard mining to CRL joint learning (ablation) | CRL · X-Domain | mean accuracy (%): 78.89 -> 80.42 | Improved |
| use class-level instead of instance-level hard mining with relative loss (ablation) | CRL · X-Domain | gain in mean class-balanced accuracy (%): 6.46 -> 6.89 | Improved |
| use relative comparison loss instead of distribution loss (ablation) | CRL class-level · CelebA | gain in mean class-balanced accuracy (%): 4.90 -> 6.32 | Improved |
| apply CRL to all classes instead of only minority classes (ablation) | CRL · X-Domain | mean accuracy (%): 80.42 -> 81.30 | Mixed |
| raise minority class criterion rho from 10% to 50% (ablation) | CRL · X-Domain | mean accuracy (%): 77.82 -> 80.42 | Improved |
| use top-kappa hard mining with kappa=1 (hardest only) (ablation) | CRL · X-Domain | accuracy | Hurt |
| use Down-Sampling instead of CE baseline | MTCT · X-Domain | mean class-balanced accuracy (%): 73.53 -> 26.29 | Hurt |
Closest papers by meaning
- Gradient Reweighting: Towards Imbalanced Class-Incremental Learning (CVPR 2024 · similarity 0.89)
- Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · similarity 0.87)
- Learning under Localized Minority Imbalance (NeurIPS 2026 · similarity 0.87)
- Revive Re-weighting in Imbalanced Learning by Density Ratio Estimation (NeurIPS 2024 · similarity 0.86)
- Synthetic Tabular Data Generation for Imbalanced Classification: The Surprising Effectiveness of an Overlap Class (AAAI 2025 · similarity 0.86)
- Rebalancing Multi-Label Class-Incremental Learning (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.
