Research map / Label-efficient and robust learning / OOD and anomaly detection
A Statistical Framework for Efficient Out of Distribution Detection in Deep Neural Networks
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 |
|---|---|---|---|
| Use MaSF instead of GRAM | OOD detection on the benchmark across DenseNet and ResNet · OOD benchmark | mTPR: 94.9 -> 96.4 | Helped |
| Use Max spatial pooling rather than Mean (ablation) | MaSF hierarchical detector · OOD benchmark | mTPR: 91.1 -> 96.4 | Helped |
| Use Simes rather than Fisher for channel reduction (ablation) | Max-spatial, Fisher-layer detector · OOD benchmark | mTPR: 95.3 -> 96.4 | Helped |
| Use Fisher rather than Simes for layer reduction (ablation) | Max-spatial, Simes-channel detector · OOD benchmark | mTPR: 95.3 -> 96.4 | Helped |
| Aggregate evidence from all layers rather than using only the AP input (ablation) | MaSF · OOD benchmark | mTPR: 64.7 -> 96.4 | Helped |
| Randomly retain only 10% of channels per layer (ablation) | MaSF · OOD benchmark | mTPR: 96.4 -> 95.2 | Hurt |
| Randomly retain only 5% of channels per layer (ablation) | MaSF · OOD benchmark | mTPR: 96.4 -> 94.5 | Hurt |
| Use Fisher instead of Simes for channel reduction (ablation) | Max-spatial detector with Fisher layer reduction · OOD benchmark | mTPR: 96.4 -> 95.3 | Hurt |
| Use Simes instead of Fisher for layer reduction (ablation) | Max-spatial detector with Simes channel reduction · OOD benchmark | mTPR: 96.4 -> 95.3 | Hurt |
| Use Mean spatial pooling instead of Max (ablation) | MaSF · OOD benchmark | mTPR: 96.4 -> 91.1 | Hurt |
| Randomly retain 75% of channels per layer (ablation) | MaSF · OOD benchmark | mTPR: 96.4 -> 96.3 | Hurt |
| Randomly retain 50% of channels per layer (ablation) | MaSF · OOD benchmark | mTPR: 96.4 -> 96.1 | Hurt |
Closest papers by meaning
- Out-of-Distribution Detection based on In-Distribution Data Patterns Memorization with Modern Hopfield Energy (ICLR 2023 · similarity 0.92)
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · similarity 0.90)
- OODD: Test-time Out-of-Distribution Detection with Dynamic Dictionary (CVPR 2025 · similarity 0.90)
- Test-Time Linear Out-of-Distribution Detection (CVPR 2024 · similarity 0.89)
- Exploiting Discrepancy in Feature Statistic for Out-of-Distribution Detection (AAAI 2024 · similarity 0.89)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 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.