Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Quantification of Uncertainty with Adversarial Models
Kajetan Schweighofer, Lukas Aichberger, Mykyta Ielanskyi, Günter Klambauer, Sepp Hochreiter
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 QUAM instead of Deep Ensembles epistemic uncertainty | uncertainty estimation · MNIST vs FMNIST | AUROC: .988 -> .994 | Improved |
| use QUAM instead of Deep Ensembles | uncertainty estimation · MNIST vs EMNIST | AUROC: .924 -> .937 | Improved |
| use QUAM instead of Deep Ensembles | uncertainty estimation · MNIST vs OMNIGLOT | AUROC: .983 -> .992 | Improved |
| use cSG-HMC, Laplace or MC dropout epistemic uncertainty instead of reference aleatoric uncertainty | reference LeNet · MNIST vs FMNIST | AUROC: .986 -> .977 | Hurt |
| use QUAM instead of Deep Ensembles (all) | EfficientNet · ImageNet-O | AUROC: .709 -> .753 | Improved |
| use QUAM instead of Deep Ensembles (all) | EfficientNet · ImageNet-A | AUROC: .874 -> .872 | No clear effect |
| use QUAM instead of reference aleatoric uncertainty | EfficientNet · ImageNet-1K misclassification | AUROC: .867 -> .904 | Improved |
| use QUAM instead of reference aleatoric uncertainty | EfficientNet · ImageNet-1K selective prediction | AUC: .958 -> .969 | Improved |
| use DE (LL) last-layer ensemble instead of reference aleatoric uncertainty | EfficientNet · ImageNet-O | AUROC: .626 -> .562 | Hurt |
| use MC dropout instead of DE (all) | EfficientNet · ImageNet-O | AUROC: .709 -> .680 | Mixed |
| restrict adversarial model search to top N classes under matched compute (ablation) | QUAM · ImageNet-O/A | AUROC | Improved |
| targeted per-class search instead of untargeted search (ablation) | QUAM · implementation | predictive diversity | Improved |
Closest papers by meaning
- Sampling-Free Epistemic Uncertainty Estimation Using Approximated Variance Propagation (ICCV 2019 · similarity 0.88)
- Epistemic Uncertainty Quantification For Pre-Trained Neural Networks (CVPR 2024 · similarity 0.88)
- Single-Model Uncertainties for Deep Learning (NeurIPS 2019 · similarity 0.88)
- Are Uncertainty Quantification Capabilities of Evidential Deep Learning a Mirage? (NeurIPS 2024 · similarity 0.87)
- Is Epistemic Uncertainty Faithfully Represented by Evidential Deep Learning Methods? (ICML 2024 · similarity 0.86)
- Benchmarking Uncertainty Disentanglement: Specialized Uncertainties for Specialized Tasks (NeurIPS 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.
