Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
All You Need is a Good Functional Prior for Bayesian Deep Learning
Ba-Hien Tran, Simone Rossi, Dimitrios Milios, Maurizio Filippone
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 a GP-induced hierarchical weight prior instead of a fixed hierarchical prior (ablation) | Bayesian CNN · CIFAR10, LeNet5 | accuracy: 75.22 -> 76.51 | Improved |
| Use a GP-induced hierarchical weight prior instead of a fixed hierarchical prior (ablation) | Bayesian CNN · CIFAR10, PreResNet20 | accuracy: 87.26 -> 88.20 | Improved |
| Use a GP-induced hierarchical weight prior instead of a fixed hierarchical prior (ablation) | Bayesian CNN · CIFAR10, VGG16 | accuracy: 86.03 -> 87.03 | Improved |
| Use a GP-induced Gaussian weight prior instead of a fixed Gaussian prior (ablation) | Bayesian CNN · CIFAR10, LeNet5 | accuracy: 74.65 -> 75.15 | Improved |
| Use a GP-induced Gaussian weight prior instead of a fixed Gaussian prior (ablation) | Bayesian CNN · CIFAR10, PreResNet20 | accuracy: 85.34 -> 86.86 | Improved |
| Use a GP-induced Gaussian weight prior instead of a fixed Gaussian prior (ablation) | Bayesian CNN · CIFAR10, VGG16 | accuracy: 81.47 -> 83.34 | Improved |
| Apply temperature scaling to the fixed Gaussian posterior | Bayesian CNN · CIFAR10, LeNet5 | accuracy: 74.65 -> 74.08 | Hurt |
| Apply temperature scaling to the fixed Gaussian posterior | Bayesian CNN · CIFAR10, PreResNet20 | accuracy: 85.34 -> 87.70 | Improved |
| Apply temperature scaling to the fixed Gaussian posterior | Bayesian CNN · CIFAR10, VGG16 | accuracy: 81.47 -> 82.25 | Improved |
| Match the neural-network functional prior to a target GP prior rather than use standard parameter priors | Bayesian neural networks · UCI regression | test NLL | Improved |
| Use a GP-induced hierarchical prior rather than a fixed or GP-induced Gaussian prior | Bayesian neural networks · UCI classification | test NLL | Improved |
| Use Wasserstein optimization rather than KL-based prior matching | BNN functional-prior matching · UCI datasets | squared MMD | Improved |
Closest papers by meaning
- Practical Deep Learning with Bayesian Principles (NeurIPS 2019 · similarity 0.85)
- Exact marginal prior distributions of finite Bayesian neural networks (NeurIPS 2021 · similarity 0.84)
- Position: Bayesian Deep Learning is Needed in the Age of Large-Scale AI (ICML 2024 · similarity 0.84)
- FUNCTIONAL VARIATIONAL BAYESIAN NEURAL NETWORKS (ICLR 2019 · similarity 0.84)
- Improving Generalization with Flat Hilbert Bayesian Inference (ICML 2025 · similarity 0.83)
- Bayesian Low-Rank Learning (Bella): A Practical Approach to Bayesian Neural Networks (AAAI 2025 · similarity 0.83)
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.
