Research map / Label-efficient and robust learning / Recognition and long-tail
A Unified Perspective for Loss-Oriented Imbalanced Learning via Localization
Zitai Wang, Qianqian Xu, Zhiyong Yang, Zhi Qiang Xu, Linchao Zhang, 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 ADRW re-weighting to CVS (ablation) | CVS · CIFAR-100 LT (rho=100, protocol a) | balanced accuracy: 43.5 -> 44.4 | Improved |
| add ADRW re-weighting to CVS (ablation) | CVS · CIFAR-100 LT (rho=100, all classes, PEFT ViT) | balanced accuracy: 79.9 -> 80.3 | Improved |
| add ADRW re-weighting to CVS (ablation) | CVS · CIFAR-10 LT (rho=100, protocol a) | balanced accuracy: 81.3 -> 81.5 | Improved |
| add ADRW re-weighting to CE (ablation) | CE · CIFAR-10 LT (rho=100, protocol a) | balanced accuracy: 71.5 -> 78.6 | Improved |
| add ADRW re-weighting to CE (ablation) | CE · CIFAR-100 LT (rho=100, protocol a) | balanced accuracy: 38.3 -> 41.8 | Improved |
| add DRW re-weighting to VS (ablation) | VS · CIFAR-100 LT (rho=100, protocol a) | balanced accuracy: 41.8 -> 41.3 | Hurt |
| add DRW to VS (ablation) | VS · CIFAR-10 LT (rho=100, protocol a) | balanced accuracy: 78.8 -> 80.1 | Improved |
| use CVS loss instead of VS+TLA+ADRW (conference version) | VS+TLA+ADRW · CIFAR-100 LT (rho=100, protocol b tuned wd) | balanced accuracy: 46.3 -> 48.2 | Improved |
| tune weight decay (protocol b instead of a) (ablation) | VS · CIFAR-100 LT (rho=100) | balanced accuracy: 41.8 -> 46.3 | Improved |
| use CVS+ADRW instead of VS | VS · CIFAR-100 LT (rho=100, all classes, FFT ViT) | balanced accuracy: 79.1 -> 80.0 | Improved |
| use CVS instead of VS (ablation) | VS · ImageNet-LT (ResNet-50) | balanced accuracy: 54.1 -> 54.7 | Improved |
| use CVS instead of Origin (ablation) | Ours-Origin · iNaturalist (ResNet-50) | balanced accuracy: 70.7 -> 71.5 | Improved |
Closest papers by meaning
- Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · similarity 0.90)
- Learning under Localized Minority Imbalance (NeurIPS 2026 · similarity 0.87)
- Improved Balanced Classification with Theoretically Grounded Loss Functions (NeurIPS 2025 · similarity 0.87)
- Dist Loss: Enhancing Regression in Few-Shot Region through Distribution Distance Constraint (ICLR 2025 · similarity 0.87)
- Gradient Reweighting: Towards Imbalanced Class-Incremental Learning (CVPR 2024 · similarity 0.86)
- Improve Representation for Imbalanced Regression through Geometric Constraints (CVPR 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.
