Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Uncertainty Quantification for Machine Learning: One Size Does Not Fit All
Paul Hofman, Yusuf Sale, Eyke Hüllermeier
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 |
|---|---|---|---|
| instantiate uncertainty measure with the task loss (zero-one/log/Brier) for rejection (ablation) | total uncertainty · Selective prediction, CovType RandomForest | AULC task loss: lowest loss-rejection curve when uncertainty loss matches task loss | Improved |
| use log-loss (mutual information) epistemic uncertainty instead of zero-one (ablation) | epistemic uncertainty · OoD detection CIFAR-10 vs SVHN, ensemble | AUROC: 0.776±0.007 -> 0.872±0.005 | Improved |
| use log-loss instead of Brier epistemic uncertainty (ablation) | epistemic uncertainty · OoD detection CIFAR-10 vs CIFAR-100, ensemble | AUROC: 0.852±0.002 -> 0.860±0.001 | Improved |
| use log-loss instead of zero-one epistemic uncertainty (ablation) | epistemic uncertainty, Dropout · OoD detection CIFAR-10 vs CIFAR-100 | AUROC: 0.702±0.001 -> 0.829±0.000 | Improved |
| use zero-one epistemic uncertainty for query selection (ablation) | epistemic uncertainty · Active learning MNIST, TissueMNIST, BloodMNIST | zero-one loss | Improved |
| use only aleatoric or epistemic component for rejection instead of total uncertainty | selective prediction · Selective prediction | loss-rejection | Hurt |
Closest papers by meaning
- Second-Order Uncertainty Quantification: A Distance-Based Approach (ICML 2024 · similarity 0.91)
- Rethinking Aleatoric and Epistemic Uncertainty (ICML 2025 · similarity 0.89)
- Is Epistemic Uncertainty Faithfully Represented by Evidential Deep Learning Methods? (ICML 2024 · similarity 0.88)
- From Risk to Uncertainty: Generating Predictive Uncertainty Measures via Bayesian Estimation (ICLR 2025 · similarity 0.88)
- Evaluating Epistemic Uncertainty: Beyond OOD Detection and Active Learning (NeurIPS 2026 · similarity 0.87)
- Are Uncertainty Quantification Capabilities of Evidential Deep Learning a Mirage? (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.
