Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Laplace Redux - Effortless 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 |
|---|---|---|---|
| apply LA post-hoc on last layer instead of online on all weights (ablation) | Laplace approximation · CIFAR-10/MNIST/FashionMNIST | in/out-of-distribution performance | Helped |
| use KFAC covariance instead of diagonal (ablation) | Laplace approximation · image classification benchmarks | performance | Helped |
| use empirical Fisher instead of GGN Hessian approximation (ablation) | Laplace approximation · OOD detection vs in-distribution | performance | Mixed |
| use probit predictive instead of Laplace bridge (ablation) | last-layer LA · image classification | performance | Helped |
| use full empirical Fisher with probit (LA*) instead of default LA (ablation) | last-layer LA · MNIST OOD detection | Confidence: 67.5 -> 56.1 | Helped |
| LA* instead of MAP | MNIST classifier · OOD detection | Confidence: 75.0 -> 56.1 | Helped |
| LA instead of deep ensembles | CIFAR-10 · OOD detection AUROC | AUROC: 94.0 (DE) vs 92.2 (LA) | Hurt |
| LA instead of MAP | CIFAR-10 · OOD detection AUROC | AUROC: 92.1 -> 92.2 | No clear effect |
Closest papers by meaning
- FSP-Laplace: Function-Space Priors for the Laplace Approximation in Bayesian Deep Learning (NeurIPS 2024 · similarity 0.92)
- Variational Linearized Laplace Approximation for Bayesian Deep Learning (ICML 2024 · similarity 0.89)
- Laplace Sample Information: Data Informativeness Through a Bayesian Lens (ICLR 2025 · similarity 0.87)
- Scalable Bayesian Learning with posteriors (ICLR 2025 · similarity 0.86)
- Reparameterization invariance in approximate Bayesian inference (NeurIPS 2024 · similarity 0.86)
- Bayesian Low-Rank Learning (Bella): A Practical Approach to Bayesian Neural Networks (AAAI 2025 · 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.