Research map / Label-efficient and robust learning / OOD and anomaly detection
RankFeat: Rank-1 Feature Removal for Out-of-distribution Detection
Yue Song, Nicu Sebe, Wei Wang
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 |
|---|---|---|---|
| remove the dominant rank-1 feature matrix | ResNetv2-101 with RankFeat · ImageNet-1k OOD suite (average) | FPR95: 57.66 -> 36.80 | Improved |
| combine Block 3 and Block 4 scores by averaging logits (ablation) | RankFeat · ImageNet-1k OOD suite (average) | FPR95: 39.69 -> 36.80 | Improved |
| keep only the rank-1 matrix rather than remove it (ablation) | RankFeat · ImageNet-1k OOD suite (average) | FPR95: 39.69 -> 58.49 | Hurt |
| remove rank-2 instead of rank-1 (ablation) | RankFeat · ImageNet-1k OOD suite (average) | FPR95: 39.69 -> 57.92 | Hurt |
| remove rank-3 instead of rank-1 (ablation) | RankFeat · ImageNet-1k OOD suite (average) | FPR95: 39.69 -> 61.94 | Hurt |
| apply rank-1 removal at shallow Block 1 rather than Block 4 (ablation) | RankFeat · ImageNet-1k OOD suite (average) | FPR95: 39.69 -> 76.65 | Hurt |
| apply rank-1 removal at shallow Block 2 rather than Block 4 (ablation) | RankFeat · ImageNet-1k OOD suite (average) | FPR95: 39.69 -> 73.23 | Hurt |
| use Block 3 rather than Block 4 features (ablation) | RankFeat · ImageNet-1k OOD suite (average) | AUROC: 87.84 -> 90.80 | Improved |
| use 20 power iterations to approximate the dominant singular component (ablation) | RankFeat · ImageNet-1k OOD suite (average) | FPR95: 39.69 -> 39.75 | No clear effect |
| reduce the power iteration count to 5 (ablation) | RankFeat · ImageNet-1k OOD suite (average) | FPR95: 39.69 -> 41.36 | Hurt |
| remove the rank-1 component from OOD features (ablation) | ID/OOD feature matrices at Block 4 · ImageNet-1k feature statistics | KL divergence to fitted MP distribution (OOD): 18.24 -> 15.79 | Improved |
Later papers that compared with this method
1 later accepted paper reports RankFeat as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- DisCoPatch: Taming Adversarially-driven Batch Statistics for Improved Out-of-Distribution Detection (ICCV 2025 · better on 6 of 6 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ReAct — ReAct: Out-of-distribution Detection With Rectified Activations (NeurIPS 2021) · better on 2 of 2 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
- Learning to Shape In-distribution Feature Space for Out-of-distribution Detection (NeurIPS 2024 · similarity 0.88)
- Test-Time Linear Out-of-Distribution Detection (CVPR 2024 · similarity 0.88)
- EAT: Towards Long-Tailed Out-of-Distribution Detection (AAAI 2024 · similarity 0.87)
- Long-Tailed Out-of-Distribution Detection via Normalized Outlier Distribution Adaptation (NeurIPS 2024 · similarity 0.87)
- Exploiting Discrepancy in Feature Statistic for Out-of-Distribution Detection (AAAI 2024 · similarity 0.87)
- The Best of Both Worlds: On the Dilemma of Out-of-distribution Detection (NeurIPS 2024 · similarity 0.87)
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.
