Research map / Label-efficient and robust learning / Recognition and long-tail
A Unified Generalization Analysis of Re-Weighting and Logit-Adjustment for Imbalanced Learning
Zitai Wang, Qianqian Xu, Zhiyong Yang, Yuan He, Xiaochun Cao, Qingming Huang
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 DRW re-weighting to CE (ablation) | CE · CIFAR-10 LT rho=100 | balanced accuracy: 71.5 -> 75.8 | Improved |
| replace DRW with adaptive ADRW re-weighting (ablation) | CE+DRW · CIFAR-10 LT rho=100 | balanced accuracy: 75.8 -> 78.6 | Improved |
| replace DRW with ADRW (ablation) | LDAM+DRW · CIFAR-10 LT rho=100 | balanced accuracy: 77.7 -> 79.1 | Improved |
| add DRW to VS loss (multiplicative logit adjustment) (ablation) | VS · CIFAR-100 LT rho=100 | balanced accuracy: 41.8 -> 41.3 | Hurt |
| add DRW to VS loss (ablation) | VS · CIFAR-100 Step rho=10 | balanced accuracy: 59.9 -> 58.0 | Hurt |
| add DRW to VS loss (ablation) | VS · CIFAR-10 LT rho=100 | balanced accuracy: 78.8 -> 80.1 | Improved |
| add TLA (additive logit adjustment) to VS+DRW (ablation) | VS+DRW · CIFAR-100 LT rho=100 | balanced accuracy: 41.3 -> 43.0 | Improved |
| replace DRW with ADRW in VS+TLA (ablation) | VS+TLA · CIFAR-10 Step rho=100 | balanced accuracy: 80.0 -> 80.9 | Improved |
| add SAM to VS+TLA+ADRW (ablation) | VS+TLA+ADRW · CIFAR-10 LT rho=100 | balanced accuracy: 81.1 -> 83.6 | Improved |
| add SAM to VS+TLA+ADRW (ablation) | VS+TLA+ADRW · CIFAR-100 LT rho=100 | balanced accuracy: 43.4 -> 46.4 | Improved |
| use CE+ADRW instead of LDAM+ADRW at rho=10 (ablation) | CE+ADRW · CIFAR-100 LT rho=10 | balanced accuracy: 58.0 -> 58.3 | Mixed |
| use class-balanced (CB) loss instead of DRW (ablation) | CE · ImageNet-LT | balanced accuracy (All): 44.4 -> 33.2 | Hurt |
Closest papers by meaning
- Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · similarity 0.89)
- Revive Re-weighting in Imbalanced Learning by Density Ratio Estimation (NeurIPS 2024 · similarity 0.87)
- Gradient Reweighting: Towards Imbalanced Class-Incremental Learning (CVPR 2024 · similarity 0.86)
- Class and Attribute-Aware Logit Adjustment for Generalized Long-Tail Learning (AAAI 2025 · similarity 0.86)
- Improved Balanced Classification with Theoretically Grounded Loss Functions (NeurIPS 2025 · similarity 0.85)
- Restoring balance: principled under/oversampling of data for optimal classification (ICML 2024 · similarity 0.84)
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.
