Research map / Label-efficient and robust learning / OOD and anomaly detection
Meta Learning Low Rank Covariance Factors for Energy Based Deterministic Uncertainty
Jeffrey Ryan Willette, Hae Beom Lee, Juho Lee, Sung Ju Hwang
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 |
|---|---|---|---|
| Use meta-learned diagonal or low-rank class covariances with scaled-energy inference | few-shot classifiers · Omniglot OOD classes, 5-way 5-shot | ECE: 48.01±0.76 -> 33.19±0.94 | Improved |
| Use meta-learned diagonal or low-rank class covariances with scaled-energy inference | few-shot classifiers · Omniglot OOD classes, 20-way 1-shot | ECE: 35.62±0.49 -> 49.28±1.21 | Hurt |
| Use meta-learned diagonal or low-rank class covariances with scaled-energy inference | few-shot classifiers · MiniImageNet OOD classes, 5-way 1-shot | ECE: 29.86±0.73 -> 17.19±1.80 | Improved |
| Use meta-learned diagonal covariance factors | few-shot classifiers · Omniglot-C, 5-way 5-shot | ECE: 24.85±13.12 -> 5.87±2.58 | Improved |
| Use meta-learned diagonal covariance factors | few-shot classifiers · MiniImageNet-C, 5-way 1-shot | ECE: 12.32±3.84 -> 8.13±3.72 | Improved |
| Use meta-learned diagonal covariance factors | few-shot classifiers · Meta-moons OOD | ECE: 45.74±4.19 -> 15.66±2.65 | Improved |
| Use meta-learned diagonal covariance factors | few-shot classifiers · Meta-Gaussians OOD | ECE: 62.60±6.57 -> 42.55±0.98 | Improved |
| Use rank-1 covariance factors instead of diagonal covariance (ablation) | Ours · Omniglot-C, 5-way 5-shot | ECE: 5.87±2.58 -> 6.10±2.87 | Hurt |
| Use rank-2 covariance factors instead of diagonal covariance (ablation) | Ours · Omniglot-C, 5-way 5-shot | ECE: 5.87±2.58 -> 6.62±3.09 | Hurt |
| Use rank-4 covariance factors instead of diagonal covariance (ablation) | Ours · Omniglot-C, 5-way 5-shot | ECE: 5.87±2.58 -> 7.10±3.40 | Hurt |
| Use rank-8 covariance factors instead of diagonal covariance (ablation) | Ours · Omniglot-C, 5-way 5-shot | ECE: 5.87±2.58 -> 7.05±3.42 | Hurt |
| Use rank-1 covariance factors instead of diagonal covariance (ablation) | Ours · MiniImageNet-C, 5-way 1-shot | ECE: 8.13±3.72 -> 9.02±3.83 | Hurt |
Closest papers by meaning
- Secure Out-of-Distribution Task Generalization with Energy-Based Models (NeurIPS 2023 · similarity 0.87)
- Post-hoc Uncertainty Learning Using a Dirichlet Meta-Model (AAAI 2023 · similarity 0.85)
- Adaptive Uncertainty Estimation via High-Dimensional Testing on Latent Representations (NeurIPS 2023 · similarity 0.85)
- A Distribution-dependent Analysis of Meta Learning (ICML 2021 · similarity 0.85)
- Meta-Learning Reliable Priors in the Function Space (NeurIPS 2021 · similarity 0.85)
- MetaModulation: Learning Variational Feature Hierarchies for Few-Shot Learning with Fewer Tasks (ICML 2023 · similarity 0.85)
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.
