Research map / Optimization / Optimizers
Iterative Robust Satisficing: Minimizing Performance Degradation Under Distribution Shift
Enes Ağırman, Artun Saday, Cem Tekin
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 |
|---|---|---|---|
| use IRS instead of ERM | WideResNet-28-10 · CIFAR-10-LT (IF=100) | worst-class accuracy: 0.088 -> 0.174 | Improved |
| use IRS instead of ERM | WideResNet-28-10 · CIFAR-10-LT (IF=100) | tail-class accuracy: 0.208 -> 0.285 | Improved |
| use IRS instead of ERM | WideResNet-28-10 · CIFAR-10-LT (IF=20) | worst-class accuracy: 0.470 -> 0.553 | Improved |
| use IRS instead of ERM on balanced data | WideResNet-28-10 · CIFAR-10-LT (IF=1) | average accuracy: 0.890 -> 0.885 | No clear effect |
| use IRS instead of KL-RS | WideResNet-28-10 · CIFAR-10-LT (IF=50) | worst-class accuracy: 0.000 -> 0.388 | Improved |
| use IRS instead of ERM | ResNet-50 · Waterbirds | worst-group accuracy: 63.08 -> 88.01 | Improved |
| use IRS instead of GroupDRO | ResNet-50 · Waterbirds | worst-group accuracy: 85.51 -> 88.01 | Improved |
| use IRS instead of GroupDRO | ResNet-50 · Waterbirds | overall accuracy: 92.13 -> 91.77 | No clear effect |
| use IRS instead of ERM | MLP/regressor · Bike Sharing | test MSE: 1.009 -> 0.815 | Improved |
| use IRS instead of ERM | MLP/regressor · Concrete Strength | test MSE: 1.316 -> 1.116 | Improved |
| use IRS (skips satisfied minibatches) instead of ERM | ResNet-50 · Waterbirds | wall-clock training time: 15:45.32 -> 12:29.94 | Improved |
| use SAM instead of ERM | WideResNet-28-10 · CIFAR-10-LT (IF=100) | worst-class accuracy: 0.088 -> 0.000 | Hurt |
Closest papers by meaning
- Statistical Properties of Robust Satisficing (ICML 2024 · similarity 0.90)
- Robust Training of Neural Networks at Arbitrary Precision and Sparsity (ICLR 2026 · similarity 0.86)
- Robust Bayesian Satisficing (NeurIPS 2023 · similarity 0.86)
- Distributionally Robust Optimization with Bias and Variance Reduction (ICLR 2024 · similarity 0.85)
- Learning Neural Networks with Distribution Shift: Efficiently Certifiable Guarantees (ICLR 2025 · similarity 0.85)
- Towards Robust Out-of-Distribution Generalization Bounds via Sharpness (ICLR 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.
