Research map / Label-efficient and robust learning / OOD and anomaly detection
Revisiting Deep Ensemble for Out-of-Distribution Detection: A Loss Landscape Perspective
Kun Fang, Qinghua Tao, Xiaolin Huang, Jie Yang
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 |
|---|---|---|---|
| ensemble 4 independent modes instead of one | RankFeat · ImageNet-1K vs iNaturalist (DenseNet121) | FPR95: 63.06 -> 40.92 | Improved |
| ensemble 4 independent modes instead of one | MSP · ImageNet-1K vs iNaturalist (DenseNet121) | FPR95: 58.38 -> 53.42 | Improved |
| ensemble 4 independent modes instead of one | Energy · ImageNet-1K vs iNaturalist (DenseNet121) | FPR95: 55.30 -> 49.23 | Improved |
| ensemble 4 independent modes instead of one | Mahalanobis · ImageNet-1K vs iNaturalist (DenseNet121) | FPR95: 91.04 -> 81.49 | Improved |
| ensemble 4 independent modes instead of one | GradNorm · ImageNet-1K vs iNaturalist (DenseNet121) | FPR95: 71.53 -> 67.22 | Improved |
| ensemble modes of the same trajectory (perturbed subspace) instead of independent modes (ablation) | Energy mode ensemble · ImageNet-1K and CIFAR10 | InD accuracy and OoD FPR | Hurt |
| ensemble 3 DenseNet121 instead of one ResNet50 of similar parameter count | RankFeat · ImageNet-1K vs iNaturalist | FPR95: 87.91 -> 44.81 | Improved |
| ensemble 3 DenseNet121 instead of one ResNet50 of similar parameter count | GradNorm · ImageNet-1K vs iNaturalist | FPR95: 56.87 -> 67.11 | Hurt |
| replace ZODE p-value ensemble with logit/feature mode ensemble (ablation) | Energy k=4 · ImageNet-1K (DenseNet121) | average FPR95: 56.79 -> 50.59 | Improved |
| replace ZODE p-value ensemble with logit/feature mode ensemble (ablation) | Energy k=4 · CIFAR10 (ResNet18) | average FPR95: 47.76 -> 40.18 | Improved |
| ensemble RN50 and DN121 modes of different structures | Energy · ImageNet-1K vs iNaturalist (Energy, mixed structures) | FPR95: 52.86 -> 50.41 | Improved |
| ensemble RN18 and WRN28X10 modes of different structures | Energy · CIFAR10 vs SVHN (Energy, mixed structures) | FPR95: 21.94 -> 27.49 | Mixed |
Closest papers by meaning
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · similarity 0.90)
- Detecting Out-of-Distribution Through the Lens of Neural Collapse (CVPR 2025 · similarity 0.90)
- A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts (NeurIPS 2026 · similarity 0.90)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.90)
- Mysteries of the Deep: Role of Intermediate Representations in Out of Distribution Detection (NeurIPS 2025 · 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.
