Research map / Label-efficient and robust learning / Recognition and long-tail
Boosting Open Set Recognition Performance through Modulated Representation Learning
Amit Kumar Kundu, Vaishnavi S Patil, Joseph JaJa
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 |
|---|---|---|---|
| increase temperature during training (low to high) instead of constant (ablation) | CE loss · CUB Hard split | AUROC: 74.98 -> 78.08 | Improved |
| decrease temperature during training (linear decrease or M-CosSch) instead of increasing (ablation) | CE loss · CUB | accuracy: 84.43 -> 81.64 | Hurt |
| random temperature each epoch (ablation) | CE loss · TinyImageNet | AUROC: 82.85 -> 78.37 | Hurt |
| add M-NegCosSch to SupCon loss | SupCon · CUB | accuracy: 83.43 -> 85.3 | Improved |
| add NegCosSch to ARPL loss | ARPL · SCars | accuracy: 97.37 -> 97.29 | No clear effect |
| add NegCosSch to CE+LS | CE with label smoothing · TinyImageNet | accuracy: 84.55 -> 84.24 | Mixed |
| add NegCosSch on a pretrained Tiny ViT | CE loss · CUB | accuracy: 90.83 -> 91.2 | Improved |
| vary period P (100 vs 200) of NegCosSch (ablation) | SupCon NegCosSch · TinyImageNet | AUROC: 82.91 -> 83.09 | No clear effect |
| increase delay k from 0 to 1 in generalized cosine schedule (ablation) | SupCon GCosSch · TinyImageNet | AUROC: 82.87 -> 83.09 | Improved |
| add NegCosSch to ProtoCon loss | ProtoCon · TinyImageNet | OSCR: 70.53 -> 71.12 | Improved |
| add NegCosSch on CIFAR benchmarks | SupCon · CIFAR10, CIFAR+10, CIFAR+50 | AUROC: 94.04 -> 94.10 | No clear effect |
| train with more known classes while using NegCosSch (ablation) | CE NegCosSch · SSBs | improvement over best constant | Improved |
Closest papers by meaning
- Open-Set Recognition: A Good Closed-Set Classifier is All You Need (ICLR 2022 · similarity 0.91)
- From Coarse to Fine-Grained Open-Set Recognition (CVPR 2024 · similarity 0.90)
- Exploring Diverse Representations for Open Set Recognition (AAAI 2024 · similarity 0.89)
- Frequency Shuffling and Enhancement for Open Set Recognition (AAAI 2024 · similarity 0.88)
- All Beings Are Equal in Open Set Recognition (AAAI 2024 · similarity 0.88)
- Unlocking Better Closed-Set Alignment Based on Neural Collapse for Open-Set Recognition (AAAI 2025 · similarity 0.86)
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.
