Research map / Label-efficient and robust learning / Recognition and long-tail
PMAL: Open Set Recognition via Robust Prototype Mining
Jing Lu, Yunlu Xu, Hao Li, Zhanzhan Cheng, Yi Niu
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 PMAL rather than RPL-WRN | Open-set recognition · TinyImageNet | AUROC: 80.9 -> 83.1 | Improved |
| Use PMAL rather than RPL++ | Open-set recognition · ImageNet-LT | AUROC: 55.2 -> 71.7 | Improved |
| Add high-quality prototype selection (ablation) | PMAL prototype mining · TinyImageNet | AUROC: 78.1 -> 81.6 | Improved |
| Add diversity filtering (ablation) | PMAL prototype mining · TinyImageNet | AUROC: 80.3 -> 81.9 | Improved |
| Add both high-quality selection and diversity filtering (ablation) | PMAL prototype mining · TinyImageNet | AUROC: 78.1 -> 81.9 | Improved |
| Use point-to-set distance instead of nearest-prototype distance (ablation) | PMAL embedding optimization · TinyImageNet | AUROC: +1.2% | Improved |
| Increase prototype count from 1 to 10 (ablation) | PMAL · TinyImageNet | AUROC: 79.9 -> 83.1 | Improved |
| Vary initial model count from 2 to 6 (ablation) | PMAL prototype mining · TinyImageNet | AUROC: 83.1; 83.3; 83.2; 83.0; 83.3 | No clear effect |
| Vary the prototype threshold epsilon (ablation) | PMAL prototype mining · TinyImageNet | AUROC: 73.6; 78.1; 82.6; 83.1; 81.2 | Mixed |
| Vary the margin delta (ablation) | PMAL embedding optimization · TinyImageNet | AUROC: 80.9; 82.8; 83.1; 82.1; 80.5 | Mixed |
| Select prototypes using embedding topology robustness rather than probability (ablation) | PMAL prototype mining · TinyImageNet | AUROC: 79.3 -> 83.1 | Improved |
| Select prototypes using embedding topology robustness rather than deep ensembles (ablation) | PMAL prototype mining · TinyImageNet | AUROC: 80.5 -> 83.1 | Improved |
Later papers that compared with this method
1 later accepted paper reports PMAL as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Frequency Shuffling and Enhancement for Open Set Recognition (AAAI 2024 · better on 4 of 4 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Recent Advances in Open Set Recognition: A Survey (TPAMI 2020 · similarity 0.86)
- Open-Set Recognition: A Good Closed-Set Classifier is All You Need (ICLR 2022 · similarity 0.86)
- Sparse Representation-Based Open Set Recognition (TPAMI 2016 · similarity 0.86)
- Reason and Discovery: A New Paradigm for Open Set Recognition (TPAMI 2025 · similarity 0.85)
- All Beings Are Equal in Open Set Recognition (AAAI 2024 · similarity 0.84)
- Exploring Diverse Representations for Open Set Recognition (AAAI 2024 · similarity 0.83)
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.
