Research map / Label-efficient and robust learning / Recognition and long-tail
Escaping Saddle Points for Effective Generalization on Class-Imbalanced Data
Harsh Rangwani, Sumukh K Aithal, Mayank Mishra, Venkatesh Babu Radhakrishnan
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 SAM with high rho | CE + DRW · CIFAR-10 LT | accuracy: 75.5 -> 80.6 | Improved |
| add SAM with high rho | CE + DRW · CIFAR-100 LT | accuracy: 41.0 -> 44.6 | Improved |
| add SAM | LDAM + DRW · CIFAR-10 LT | accuracy: 77.5 -> 81.9 | Improved |
| add SAM | VS loss · CIFAR-10 LT | accuracy: 78.6 -> 82.4 | Improved |
| add SAM | CE + DRW · CIFAR-10 step imbalance | accuracy: 72.2 -> 79.3 | Improved |
| add SAM | LDAM + DRW · ImageNet-LT | accuracy: 49.9 -> 53.1 | Improved |
| add SAM | LDAM + DRW · iNaturalist 2018 | accuracy: 67.5 -> 70.1 | Improved |
| add SAM without re-weighting (ablation) | CE · CIFAR-10 LT | tail accuracy: 52.3 -> 51.7 | No clear effect |
| add SAM without re-weighting (ablation) | CE · CIFAR-10 LT | overall accuracy: 71.7 -> 73.1 | Improved |
| add PGD noise (ablation) | CE + DRW · CIFAR-10 LT | accuracy: 75.5 -> 77.2 | Improved |
| add LPF-SGD (ablation) | CE + DRW · CIFAR-10 LT | accuracy: 75.5 -> 78.5 | Improved |
| increase SAM rho from 0.05 to 0.8 (ablation) | CE + DRW + SAM · CIFAR-10 LT | tail accuracy: 63% -> 73% | Improved |
Closest papers by meaning
- Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · similarity 0.90)
- SSE-SAM: Balancing Head and Tail Classes Gradually Through Stage-Wise SAM (AAAI 2025 · similarity 0.88)
- Improved Balanced Classification with Theoretically Grounded Loss Functions (NeurIPS 2025 · similarity 0.86)
- Neural Collapse To Multiple Centers For Imbalanced Data (NeurIPS 2024 · similarity 0.86)
- Neural Collapse for Cross-entropy Class-Imbalanced Learning with Unconstrained ReLU Features Model (ICML 2024 · similarity 0.86)
- Gradient Reweighting: Towards Imbalanced Class-Incremental Learning (CVPR 2024 · 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.
