Research map / Label-efficient and robust learning / Recognition and long-tail
Deep Stable Learning for Out-of-Distribution Generalization
Check 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 sample weighting to decorrelate features with RFF | StableNet · PACS and VLCS, unbalanced 5:1:1 | accuracy (Avg.): 76.06 -> 79.66; 64.54 -> 67.99 | Helped |
| Use sample weighting to decorrelate features with RFF | StableNet · PACS, unbalanced + flexible 5:1:1 | accuracy: 42.65 -> 45.14 | Helped |
| Use sample weighting to decorrelate features with RFF | StableNet · VLCS, unbalanced + flexible 5:1:1 | accuracy: 78.96 -> 79.15 | Helped |
| Use sample weighting to decorrelate features with RFF | StableNet · NICO, unbalanced + flexible 5:1:1 | accuracy: 57.59 -> 59.76 | Helped |
| Use sample weighting to decorrelate features with RFF | StableNet · MNIST-M, unbalanced + flexible + adversarial | accuracy (Avg.): 86.11 -> 87.94 | Helped |
| Use sample weighting to decorrelate features with RFF | StableNet · PACS, classic | accuracy (Avg.): 85.15 -> 84.69 | Hurt |
| Use sample weighting to decorrelate features with RFF | StableNet · VLCS, classic | accuracy (Avg.): 76.18 -> 77.65 | Helped |
| Add nonlinear RFF decorrelation rather than linear-only decorrelation (ablation) | StableNet · NICO | accuracy | Helped |
| Increase the RFF sampling dimension (ablation) | StableNet · NICO | accuracy | Helped |
| Reduce the RFF sampling dimension (ablation) | StableNet · NICO | computational cost | Helped |
| Use presaved global features and weights rather than no saved features (ablation) | StableNet · NICO | accuracy | Helped |
| Use presaved global features and weights rather than no saved features (ablation) | StableNet · NICO | variance | Helped |
Closest papers by meaning
- OoD-Bench: Quantifying and Understanding Two Dimensions of Out-of-Distribution Generalization (CVPR 2022 · similarity 0.88)
- Uncertainty Modeling for Out-of-Distribution Generalization (ICLR 2022 · similarity 0.88)
- Distribution Free Domain Generalization (ICML 2023 · similarity 0.87)
- Assaying Out-Of-Distribution Generalization in Transfer Learning (NeurIPS 2022 · similarity 0.87)
- Feature Contamination: Neural Networks Learn Uncorrelated Features and Fail to Generalize (ICML 2024 · similarity 0.86)
- Ensemble Pruning for Out-of-distribution Generalization (ICML 2024 · 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.