Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
A Simple Baseline for Bayesian Uncertainty in 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 SWAG Bayesian model averaging rather than SWA | PreResNet-164 · CIFAR-100 | NLL: 0.7370±0.0265 -> 0.6595±0.0019 | Helped |
| Use low-rank plus diagonal covariance rather than diagonal covariance alone (ablation) | SWAG · CIFAR-100 | NLL: 0.6150±0.0029 -> 0.6078±0.0006 | Helped |
| Use low-rank plus diagonal covariance rather than diagonal covariance alone (ablation) | SWAG · CIFAR-10 | NLL: 0.2200±0.0078 -> 0.2016±0.0031 | Helped |
| Use SWAG rather than SWA-Dropout | PreResNet-164 · CIFAR-10 | NLL: 0.1270±0.0000 -> 0.1232±0.0022 | Helped |
| Use SWAG instead of SWA | LSTM · Penn Treebank test | perplexity: 56.7 -> 56.26 | Helped |
| Use SWAG instead of SWA | LSTM · WikiText-2 test | perplexity: 65.0 -> 64.1 | Helped |
| Use SWAG rather than SWA temperature scaling | VGG-16 · CIFAR-10 to STL-10 | ECE: 0.2089±0.0055 -> 0.1463±0.0075 | Helped |
| Use low-rank plus diagonal covariance rather than diagonal covariance alone (ablation) | SWAG · CIFAR-10 five-class training, held-out classes | JS-distance: 2.27 -> 3.31 | Helped |
| Use a single SWAG model rather than a 3-model SGD ensemble | PreResNet-164 · CIFAR-100 | NLL: 0.6922 -> 0.6595 | Helped |
| Use 5 independently trained SGD models instead of one SWAG model | PreResNet-164 · CIFAR-100 | NLL: 0.6595 -> 0.6478 | Helped |
Closest papers by meaning
- Sparse Uncertainty Representation in Deep Learning with Inducing Weights (NeurIPS 2021 · similarity 0.88)
- Generalization Analysis of Stochastic Weight Averaging with General Sampling (ICML 2024 · similarity 0.88)
- Variational Linearized Laplace Approximation for Bayesian Deep Learning (ICML 2024 · similarity 0.87)
- Bayesian Deep Learning and a Probabilistic Perspective of Generalization (NeurIPS 2020 · similarity 0.87)
- Sampling from Gaussian Process Posteriors using Stochastic Gradient Descent (NeurIPS 2023 · similarity 0.87)
- Uncertainty Quantification with the Empirical Neural Tangent Kernel (NeurIPS 2025 · 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.