Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Predictive Uncertainty Estimation via Prior Networks
Read the paperCheck 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 a Dirichlet Prior Network rather than a DNN for MNIST misclassification detection | MNIST classifier · MNIST valid+test misclassification detection | AUROC (Max.P): 98.0 -> 99.0 | Helped |
| Use a Dirichlet Prior Network rather than a DNN for MNIST misclassification detection | MNIST classifier · MNIST valid+test misclassification detection | AUPR (Max.P): 26.6 -> 43.6 | Helped |
| Use a Dirichlet Prior Network rather than the strongest baseline for CIFAR-10 misclassification detection | CIFAR-10 classifier · CIFAR-10 test misclassification detection | AUPR (Max.P): 48.7 -> 52.7 | Helped |
| Use a Dirichlet Prior Network rather than the strongest baseline for CIFAR-10 misclassification detection | CIFAR-10 classifier · CIFAR-10 test misclassification detection | AUROC (Max.P): 92.5 -> 92.2 | Hurt |
| Use a Dirichlet Prior Network rather than MCDP for SVHN misclassification detection | SVHN classifier · SVHN test misclassification detection | AUPR (Max.P): 46.4 -> 55.3 | Helped |
| Use a Dirichlet Prior Network rather than MCDP for SVHN misclassification detection | SVHN classifier · SVHN test misclassification detection | AUROC (Entropy): 92.2 -> 90.1 | Hurt |
| Use a Dirichlet Prior Network rather than a DNN for CIFAR-10 vs TinyImageNet OOD detection | CIFAR-10 classifier · CIFAR-10 vs TinyImageNet OOD detection | AUROC (D.Ent.): 88.7 -> 94.6 | Helped |
| Use a Dirichlet Prior Network rather than a DNN for MNIST vs OMNIGLOT OOD detection | MNIST classifier · MNIST vs OMNIGLOT OOD detection | AUROC (Max.P): 98.7 -> 100.0 | Helped |
| Use differential entropy rather than predictive entropy to detect OOD under strong input noise | MNIST DPN · MNIST vs OMNIGLOT OOD detection, noisy inputs | AUROC: 51.8 -> 99.8 | Helped |
| Add strong Gaussian noise to inputs | MNIST OOD detection · MNIST vs OMNIGLOT OOD detection | AUROC (Entropy): 100.0 -> 51.8 | Hurt |
| Add strong Gaussian noise to inputs | MNIST DPN OOD detection · MNIST vs OMNIGLOT OOD detection | AUROC (D.Ent.): 100.0 -> 99.8 | No clear effect |
| Replace entropy with differential entropy for identifying OOD inputs when classes overlap | Dirichlet Prior Network · Synthetic overlapping-class data | OOD and in-domain separability | Helped |
Closest papers by meaning
- Posterior Network: Uncertainty Estimation without OOD Samples via Density-Based Pseudo-Counts (NeurIPS 2020 · similarity 0.90)
- Adaptive Uncertainty Estimation via High-Dimensional Testing on Latent Representations (NeurIPS 2023 · similarity 0.88)
- Distribution-Free Data Uncertainty for Neural Network Regression (ICLR 2025 · similarity 0.88)
- Quantifying Uncertainty in the Presence of Distribution Shifts (NeurIPS 2025 · similarity 0.87)
- Uncertainty Quantification with the Empirical Neural Tangent Kernel (NeurIPS 2025 · similarity 0.85)
- Epistemic Uncertainty Quantification For Pre-Trained Neural Networks (CVPR 2024 · 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.