Research map / Label-efficient and robust learning / Recognition and long-tail
DRAUC: An Instance-wise Distributionally Robust AUC Optimization Framework
Siran Dai, Qianqian Xu, Zhiyong Yang, 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 |
|---|---|---|---|
| Use DRAUC-Df instead of WDRO | ResNet20 · Tiny-ImageNet-C, Dogs | AUC: 82.20 -> 84.11 | Improved |
| Use DRAUC-Da instead of DRAUC-Df (ablation) | DRAUC · CIFAR10-C, ResNet20, imbalance 0.05 | AUC: 80.18 -> 80.13 | Hurt |
| Use DRAUC-Da instead of DRAUC-Df (ablation) | DRAUC · CIFAR100-LT, ResNet32, imbalance 0.05 | AUC: 63.90 -> 64.39 | Improved |
| Use DRAUC-Da instead of DRAUC-Df (ablation) | DRAUC · CIFAR100-C, ResNet20, imbalance 0.10 | AUC: 66.25 -> 66.11 | Hurt |
| Use DRAUC-Da instead of DRAUC-Df (ablation) | DRAUC · CIFAR10-C, ResNet32, imbalance 0.20 | AUC: 90.62 -> 90.15 | Hurt |
| Apply class-wise distributional perturbation rather than distribution-free perturbation (ablation) | DRAUC · Corrupted benchmarks overall | AUC | Mixed |
| Use DRAUC rather than AUCMLoss under matched hyperparameters (ablation) | AUCMLoss · Corrupted benchmarks | AUC | Improved |
| Increase the robust diameter epsilon (ablation) | DRAUC · Relatively balanced training sets, imbalance ratio at least 0.1 | robust performance | Improved |
| Increase epsilon in highly imbalanced settings (ablation) | DRAUC · Highly imbalanced training sets | robust performance | No clear effect |
| Use a large lambda learning rate when the imbalance ratio is 0.01 (ablation) | DRAUC · Extremely imbalanced training sets | robust performance | Hurt |
| Choose an appropriate lambda learning rate around 0.1 (ablation) | DRAUC · Training sets the model can fit without difficulty | robustness: 0.1 | Improved |
| Train for robustness with DRAUC on less imbalanced data | DRAUC · Non-corrupted datasets with less extreme imbalance | AUC | Hurt |
Closest papers by meaning
- Asymptotically Unbiased Instance-wise Regularized Partial AUC Optimization: Theory and Algorithm (NeurIPS 2022 · similarity 0.87)
- When AUC meets DRO: Optimizing Partial AUC for Deep Learning with Non-Convex Convergence Guarantee (ICML 2022 · similarity 0.87)
- Incorporating Unlabeled Data into Distributionally Robust Learning (JMLR 2021 · similarity 0.87)
- AUC-Oriented Domain Adaptation: From Theory to Algorithm (TPAMI 2023 · similarity 0.86)
- Large-Scale Non-convex Stochastic Constrained Distributionally Robust Optimization (AAAI 2024 · similarity 0.85)
- Preserving AUC Fairness in Learning with Noisy Protected Groups (ICML 2025 · similarity 0.85)
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.
