Research map / Learning theory and algorithms / Prediction and decision losses
Optimizing Partial Area Under the Top-K Curve: Theory and Practice
Zitai Wang, Qianqian Xu, Zhiyong Yang, Yuan He, 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 |
|---|---|---|---|
| optimize AUTKC (Exp surrogate) instead of cross-entropy | ResNet-18 · Cifar-100 (K=3) | AUTKC: 71.81 -> 73.60 | Improved |
| optimize AUTKC (Exp surrogate) instead of cross-entropy | ResNet-18 · Tiny-imagenet-200 (K=3) | AUTKC: 81.56 -> 82.87 | Improved |
| optimize AUTKC (Exp surrogate) instead of cross-entropy | ResNet-50 features + FC · Places-365 (K=3) | AUTKC: 67.72 -> 67.95 | Improved |
| optimize AUTKC (Exp surrogate) instead of cross-entropy | ResNet-18 · Cifar-10 (K=3) | AUTKC: 92.87 -> 93.01 | Improved |
| use top-k losses instead of cross-entropy | ResNet-18 · Cifar-10 (K=3) | AUTKC: 92.87 -> 92.56 | Hurt |
| use top-k loss L3 instead of cross-entropy | ResNet-18 · Cifar-100 (K=3) | AUTKC: 71.81 -> 71.98 | Improved |
| use hinge surrogate instead of exp surrogate in AUTKC | AUTKC · Cifar-100 (K=3) | AUTKC: 73.60 -> 73.29 | Hurt |
| use exp surrogate instead of hinge surrogate in AUTKC | AUTKC · Tiny-imagenet-200 (K=3) | AUTKC: 82.20 -> 82.87 | Improved |
| use AUTKC-Hinge instead of traditional hinge loss | ResNet-18 · Cifar-100 (K=3) | AUTKC: 71.33 -> 73.29 | Improved |
| use AUTKC-Logit instead of CE | ResNet-18 · Cifar-10 (K=3) | AUTKC: 92.87 -> 93.05 | Improved |
Closest papers by meaning
- Top-k Hierarchical Classification (AAAI 2017 · similarity 0.88)
- Top-K Pairwise Ranking: Bridging the Gap Among Ranking-Based Measures for Multi-label Classification (IJCV 2024 · similarity 0.86)
- On the consistency of top-k surrogate losses (ICML 2020 · similarity 0.86)
- Loss Functions for Top-k Error: Analysis and Insights (CVPR 2016 · similarity 0.86)
- Analysis and Optimization of Loss Functions for Multiclass, Top-k, and Multilabel Classification (TPAMI 2017 · similarity 0.86)
- Cardinality-Aware Set Prediction and Top-$k$ Classification (NeurIPS 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.
