Research map / Learning theory and algorithms / Prediction and decision losses
Towards Decision-Friendly AUC: Learning Multi-Classifier with AUCµ
Read the paperCheck 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 |
|---|---|---|---|
| Optimize AUCµ with surrogate losses instead of MAUC | ResNet-18 multiclass classifier · CIFAR-100, imbalance ratio 100 | AUCµ: 92.41 -> 93.54 | Helped |
| Optimize AUCµ with a square-loss surrogate instead of the strongest listed non-AUC baseline | ResNet-18 multiclass classifier · CIFAR-100, imbalance ratio 50 | AUCµ: 93.40 -> 94.51 | Helped |
| Optimize AUCµ with a square-loss surrogate instead of the strongest listed non-AUC baseline | ResNet-18 multiclass classifier · ImageNet, imbalance ratio 50 | AUCµ: 96.71 -> 96.75 | Helped |
| Use the AUCµ method rather than MAUC-based training | ResNet-18 multiclass classifier · CIFAR-10, imbalance ratio 200 | F1-metric: 43.89 -> 50.73 | Helped |
| Learn decision thresholds using soft F1 (ablation) | Ours + Square · ImageNet, imbalance ratio 100 | F1-metric: 12.85 -> 12.90 | Helped |
| Learn decision thresholds using soft F1 (ablation) | Ours + Exp · CIFAR-100, imbalance ratio 200 | F1-metric: 21.78 -> 22.01 | No clear effect |
| Use the square-loss surrogate rather than the exponential-loss surrogate (ablation) | AUCµ training · CIFAR-10, imbalance ratio 200 | AUCµ: 91.81 vs 93.28 | Hurt |
| Increase warm-up epochs from 1 to 30 (ablation) | Proposed method on CIFAR-10, imbalance ratio 50 · CIFAR-10, imbalance ratio 50 | F1-metric | Helped |
| Vary the exponential-loss hyperparameter a across [0.5, 3.0] (ablation) | Proposed method on CIFAR-10, imbalance ratio 50 · CIFAR-10, imbalance ratio 50 | F1-metric | No clear effect |
Closest papers by meaning
- Interplay of ROC and Precision-Recall AUCs: Theoretical Limits and Practical Implications in Binary Classification (ICML 2024 · similarity 0.88)
- AUC Optimization from Multiple Unlabeled Datasets (AAAI 2024 · similarity 0.88)
- A Closer Look at AUROC and AUPRC under Class Imbalance (NeurIPS 2024 · similarity 0.87)
- Positive-unlabeled AUC Maximization under Covariate Shift (ICML 2025 · similarity 0.86)
- AUC Maximization under Positive Distribution Shift (NeurIPS 2024 · similarity 0.84)
- Pairwise AUC Optimization Needs Corrective Power: A Unified View (NeurIPS 2026 · 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.