Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Nonparametric Uncertainty Quantification for Single Deterministic Neural Network
Nikita Yurevich Kotelevskii, Aleksandr Artemenkov, Kirill Fedyanin, Fedor Noskov, Alexander Fishkov, Artem Shelmanov, Artem Vazhentsev, Aleksandr Petiushko, Maxim Panov
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 NUQ instead of DDU | ResNet-50 trained on CIFAR-100 · CIFAR-100 → SVHN | ROC-AUC: 89.6±1.6 -> 89.7±1.6 | Improved |
| Use NUQ instead of DDU | ResNet-50 trained on CIFAR-100 · CIFAR-100 → LSUN | ROC-AUC: 92.1±0.6 -> 92.3±0.6 | Improved |
| Use NUQ instead of DDU | ResNet-50 trained on CIFAR-100 · CIFAR-100 → Smooth | ROC-AUC: 97.1±3.1 -> 96.8±3.8 | Hurt |
| Use NUQ instead of SNGP | ResNet-50 trained on ImageNet · ImageNet → ImageNet-R | ROC-AUC: 85.0 -> 99.5 | Improved |
| Use NUQ instead of DDU | ResNet-50 trained on ImageNet · ImageNet → ImageNet-O | ROC-AUC: 74.1 -> 82.4 | Improved |
| Use KDE rather than GMM for the density estimate (ablation) | NUQ on ImageNet · ImageNet OOD detection | ROC-AUC | Improved |
| Use GMM rather than KDE for the density estimate (ablation) | NUQ on CIFAR-100 · CIFAR-100 OOD detection | ROC-AUC | Improved |
| Approximate the kernel estimate using nearest neighbors (ablation) | NUQ on ImageNet · ImageNet OOD detection | ROC-AUC | No clear effect |
| Use spectral normalization for the base model (ablation) | NUQ · Image and text classification tasks | uncertainty estimation performance | No clear effect |
| Use NUQ instead of MaxProb | ELECTRA text classifier · SST-2 classification with rejection | RCC-AUC | Improved |
| Use NUQ instead of DDU | ELECTRA text classifier · ROSTD and CLINC OOD detection | ROC-AUC | Improved |
Closest papers by meaning
- Uncertainty Quantification for Machine Learning: One Size Does Not Fit All (AAAI 2026 · similarity 0.88)
- A Review of Uncertainty Representation and Quantification in Neural Networks (TPAMI 2025 · similarity 0.88)
- Second-Order Uncertainty Quantification: A Distance-Based Approach (ICML 2024 · similarity 0.88)
- Epistemic Uncertainty Quantification For Pre-Trained Neural Networks (CVPR 2024 · similarity 0.87)
- Uncertainty Quantification with the Empirical Neural Tangent Kernel (NeurIPS 2025 · similarity 0.87)
- Vicinal Label Supervision for Reliable Aleatoric and Epistemic Uncertainty Estimation (NeurIPS 2025 · 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.
