Research map / Label-efficient and robust learning / Recognition and long-tail
RSG: A Simple but Effective Module for Learning Imbalanced Datasets
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 RSG to LDAM-DRW | ResNet-32 · Imbalanced CIFAR-10 long-tailed, ρ=50 | top-1 error: 18.97 -> 17.20 | Helped |
| add RSG to LDAM-DRW | ResNet-32 · Imbalanced CIFAR-10 step, ρ=50 | top-1 error: 18.67 -> 17.90 | Helped |
| add RSG to LDAM-DRW | ResNet-32 · Imbalanced CIFAR-100 long-tailed, ρ=50 | top-1 error: 53.38 -> 51.50 | Helped |
| add RSG to LDAM-DRW | ResNet-32 · Imbalanced CIFAR-100 step, ρ=50 | top-1 error: 50.97 -> 49.43 | Helped |
| add RSG to LDAM-DRS | ResNet-50 · iNaturalist 2018, 1× schedule | top-1 error: 32.73 -> 32.10 | Helped |
| use the 2× training schedule with LDAM-DRS-RSG | ResNet-50 · iNaturalist 2018 | top-1 error: 33.22 -> 29.74 | Helped |
| add RSG to LDAM-DRS | ResNet-152 · Places-LT | top-1 accuracy: 38.6 -> 39.3 | Helped |
| add RSG to LDAM-DRS | ResNeXt-50 · ImageNet-LT | top-1 accuracy: 51.4 -> 51.8 | Helped |
| add the MV-loss direction-alignment term (ablation) | RSG with ResNet-32 + LDAM-DRW · Imbalanced CIFAR-10 long-tailed, ρ=50 | top-1 error: 18.03 -> 17.20 | Helped |
| add the MV-loss vector-length preservation term (ablation) | RSG with ResNet-32 + LDAM-DRW · Imbalanced CIFAR-10 long-tailed, ρ=50 | top-1 error: 18.07 -> 17.20 | Helped |
| add the MV-loss class-relevance filtering term (ablation) | RSG with ResNet-32 + LDAM-DRW · Imbalanced CIFAR-10 long-tailed, ρ=50 | top-1 error: 17.67 -> 17.20 | Helped |
| add frequent-class displacement to rare-class samples rather than their centers (ablation) | RSG with ResNet-32 + LDAM-DRW · Imbalanced CIFAR-10 long-tailed, ρ=50 | top-1 error: 18.91 -> 17.20 | Helped |
Closest papers by meaning
- Repeated Random Sampling for Minimizing the Time-to-Accuracy of Learning (ICLR 2024 · similarity 0.84)
- Inverse Weight-Balancing for Deep Long-Tailed Learning (AAAI 2024 · similarity 0.84)
- Synthetic Tabular Data Generation for Imbalanced Classification: The Surprising Effectiveness of an Overlap Class (AAAI 2025 · similarity 0.83)
- Lightweight Dataset Pruning without Full Training via Example Difficulty and Prediction Uncertainty (ICML 2025 · similarity 0.83)
- Concentration and excess risk bounds for imbalanced classification with synthetic oversampling (NeurIPS 2025 · similarity 0.82)
- Imbalances in Neurosymbolic Learning: Characterization and Mitigating Strategies (NeurIPS 2025 · similarity 0.82)
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.