Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Gradient-Based Uncertainty Attribution for Explainable Bayesian Deep Learning
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 UA-Backprop + FullGrad attribution instead of plain training without attention | CNN uncertainty mitigation · MNIST, C10, C100, SVHN (average) | ACC (%): 48.77 -> 51.42 | Helped |
| Use UA-Backprop + FullGrad attribution instead of plain training without attention | CNN uncertainty mitigation · MNIST, C10, C100, SVHN (average) | NLL: 2.096 -> 1.968 | Helped |
| Use UA-Backprop + FullGrad rather than baseline attribution methods | Deep ensemble uncertainty maps · MNIST, C10, C100, SVHN (average) | MURR: 0.581 -> 0.607 | Helped |
| Use UA-Backprop + FullGrad rather than baseline attribution methods | Deep ensemble uncertainty maps · C10, C100, SVHN (average) | IoU: 0.289 -> 0.311 | Helped |
| Add attention to training using UA-Backprop maps (ablation) | CNN uncertainty mitigation · MNIST, C10 (average) | ACC (%): 61.20 -> 64.22 | Helped |
| Add attention to training using UA-Backprop maps (ablation) | CNN uncertainty mitigation · MNIST, C10 (average) | NLL: 1.140 -> 1.028 | Helped |
| Increase attention strength from alpha 0.2 to 0.4 (ablation) | CNN uncertainty mitigation · MNIST, C10 (average) | ACC (%): 64.22 -> 63.42 | Hurt |
| Increase attention strength from alpha 0.2 to 0.6 (ablation) | CNN uncertainty mitigation · MNIST, C10 (average) | ACC (%): 64.22 -> 63.86 | Mixed |
| Increase attention strength from alpha 0.2 to 0.8 (ablation) | CNN uncertainty mitigation · MNIST, C10 (average) | ACC (%): 64.22 -> 65.23 | Helped |
Closest papers by meaning
- Are Uncertainty Quantification Capabilities of Evidential Deep Learning a Mirage? (NeurIPS 2024 · similarity 0.88)
- Uncertainty Quantification for Deep Regression using Contextualised Normalizing Flows (NeurIPS 2025 · similarity 0.87)
- Uncertainty Quantification with the Empirical Neural Tangent Kernel (NeurIPS 2025 · similarity 0.87)
- Is Epistemic Uncertainty Faithfully Represented by Evidential Deep Learning Methods? (ICML 2024 · similarity 0.86)
- Enhancing Uncertainty Estimation and Interpretability with Bayesian Non-negative Decision Layer (ICLR 2025 · similarity 0.86)
- Epistemic Uncertainty Quantification For Pre-Trained Neural Networks (CVPR 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.