Research map / Label-efficient and robust learning / OOD and anomaly detection
Out-of-Distribution Detection with Boundary Aware Learning
Sen Pei, Xin Zhang, Bin Fan, Gaofeng Meng
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 |
|---|---|---|---|
| add BAL trivial-OOD discriminator loss L_t on top of softmax classifier (ablation) | softmax baseline · CIFAR-10 vs TinyImageNet (DenseNet-BC) | FPR95: 41.1 -> 17.9 | Improved |
| add shuffle loss L_s (ablation) | BAL (L_t) · CIFAR-10 vs TinyImageNet (DenseNet-BC) | FPR95: 17.9 -> 9.3 | Improved |
| add uniform loss L_u (ablation) | BAL (L_t) · CIFAR-10 vs TinyImageNet (DenseNet-BC) | FPR95: 17.9 -> 8.1 | Improved |
| add both shuffle and uniform loss (ablation) | BAL (L_t) · CIFAR-10 vs TinyImageNet (DenseNet-BC) | FPR95: 17.9 -> 5.0 | Improved |
| add both shuffle and uniform loss (ablation) | BAL (L_t) · CIFAR-10 vs TinyImageNet (DenseNet-BC) | AUPR_out: 96.0 -> 98.0 | Improved |
| use uniform sampling instead of Gaussian sampling of trivial OOD features (ablation) | BAL · feature dim 2, CIFAR-10 vs TinyImageNet | AUROC: 94.3 -> 96.5 | Improved |
| use uniform sampling instead of Gaussian sampling (ablation) | BAL · feature dim 512 | AUROC: 98.1 -> 98.1 | No clear effect |
| use uniform sampling instead of Gaussian sampling (ablation) | BAL · feature dim 1024 | AUROC: 98.5 -> 98.8 | Improved |
| raise lambda_c from 0.1 to 0.7 (ablation) | BAL uniform loss · CIFAR-10 vs TinyImageNet (DenseNet-BC) | AUROC: 94.8 -> 97.0 | Improved |
| raise lambda_c from 0.7 to 0.9 (ablation) | BAL uniform loss · CIFAR-10 vs TinyImageNet (DenseNet-BC) | AUROC: 97.0 -> 96.2 | Hurt |
| raise lambda_c from 0.1 to 0.9 (ablation) | BAL uniform loss · CIFAR-10 vs TinyImageNet (DenseNet-BC) | AUPR_out: 92.1 -> 97.1 | Improved |
| use BAL instead of ODIN | ODIN · Fashion-MNIST ID, MNIST & Omniglot OOD | FPR95: 33.27 -> 9.20 | Improved |
Closest papers by meaning
- Fast Decision Boundary based Out-of-Distribution Detector (ICML 2024 · similarity 0.92)
- Out-of-Distribution Detection with Prototypical Outlier Proxy (AAAI 2025 · similarity 0.90)
- BOOD: Boundary-based Out-Of-Distribution Data Generation (ICML 2025 · similarity 0.90)
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · similarity 0.89)
- Learning to Shape In-distribution Feature Space for Out-of-distribution Detection (NeurIPS 2024 · similarity 0.89)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.89)
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.
