Research map / Label-efficient and robust learning / Recognition and long-tail
Simplifying Neural Network Training Under Class Imbalance
Ravid Shwartz-Ziv, Micah Goldblum, Yucen Lily Li, C. Bayan Bruss, Andrew Gordon Wilson
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 Joint-SSL, SAM-A and M2m to M2m | M2m · CIFAR-10 ratio 0.01 | accuracy: 84.5 -> 86.0 | Improved |
| add Joint-SSL, SAM-A and M2m to ERM | ERM · CIFAR-100 ratio 0.01 | accuracy: 47.7 -> 48.9 | Improved |
| add joint self-supervised VICReg loss (ablation) | ERM · CIFAR-10 ratio 0.01 | accuracy: 84.9 -> 85.2 | Improved |
| use SAM-A alone instead of ERM (ablation) | ERM · CIFAR-10 ratio 0.02 | accuracy: 84.2 -> 83.3 | Hurt |
| reweighting instead of ERM | ERM · CIFAR-10 ratio 0.01 | accuracy: 84.9 -> 81.8 | Hurt |
| resampling instead of ERM | ERM · CIFAR-10 ratio 0.01 | accuracy: 84.9 -> 82.1 | Hurt |
| add SAM-A on top of Joint-SSL (ablation) | Joint-SSL · APTOS 2019 Blindness | accuracy: 89.7 -> 90.1 | Improved |
| add modified label smoothing on top of Joint-SSL + SAM-A (ablation) | Joint-SSL + SAM-A · SIIM-ISIC Melanoma | accuracy: 96.0 -> 96.2 | Improved |
| add modified label smoothing on top of Joint-SSL + SAM-A (ablation) | Joint-SSL + SAM-A · EuroSAT | accuracy: 99.1 -> 99.3 | Improved |
| add SAM-A and modified label smoothing to MLP | MLP · Otto | accuracy: 83.0 -> 83.2 | Improved |
| add SAM-A and modified label smoothing to MLP | MLP · CoverType | accuracy: 97.5 -> 97.6 | Improved |
| add Joint-SSL + SAM-A to ERM | ERM · CIFAR-10 neural collapse test accuracy | accuracy: 84.58 -> 84.80 | Improved |
Closest papers by meaning
- Improved Algorithm for Deep Active Learning under Imbalance via Optimal Separation (ICML 2025 · similarity 0.89)
- Gradient Reweighting: Towards Imbalanced Class-Incremental Learning (CVPR 2024 · similarity 0.87)
- Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · similarity 0.87)
- Heavy-Tailed Class Imbalance and Why Adam Outperforms Gradient Descent on Language Models (NeurIPS 2024 · similarity 0.86)
- Neural Collapse for Cross-entropy Class-Imbalanced Learning with Unconstrained ReLU Features Model (ICML 2024 · similarity 0.86)
- Inverse Weight-Balancing for Deep Long-Tailed Learning (AAAI 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.
