Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Inverse-Reference Priors for Fisher Regularization of Bayesian Neural Networks
Keunseo Kim, Eun-Yeol Ma, Jeongman Choi, Heeyoung Kim
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 the IR prior instead of the strongest competing prior | DenseNet121 BNN · CIFAR-10 | validation accuracy (%): 81.9 -> 85.5 | Improved |
| Use the IR prior instead of the strongest competing prior | ResNet18 BNN · CIFAR-10 | validation accuracy (%): 82.6 -> 84.6 | Improved |
| Use the IR prior instead of the strongest competing prior | VGG16 BNN · CIFAR-10 | validation accuracy (%): 79.4 -> 84.1 | Improved |
| Use the IR prior instead of the strongest competing prior | DenseNet121 BNN · CIFAR-100 | validation accuracy (%): 48.3 -> 52.0 | Improved |
| Use the IR prior instead of the strongest competing prior | ResNet18 BNN · CIFAR-100 | validation accuracy (%): 48.9 -> 51.1 | Improved |
| Use the IR prior instead of the strongest competing prior | VGG16 BNN · CIFAR-100 | validation accuracy (%): 48.1 -> 50.8 | Improved |
| Use the IR prior instead of the strongest competing prior | DenseNet121 BNN · SVHN | validation accuracy (%): 94.9 -> 96.1 | Improved |
| Use the IR prior instead of the strongest competing prior | ResNet18 BNN · SVHN | validation accuracy (%): 94.8 -> 95.4 | Improved |
| Use the IR prior instead of the strongest competing prior | VGG16 BNN · SVHN | validation accuracy (%): 94.3 -> 95.4 | Improved |
| Use the IR prior instead of the strongest competing prior | DenseNet BNN · CIFAR-10 adversarial examples (ϵ = 0.0001) | validation accuracy (%): 80.1 -> 84.9 | Improved |
| Use the IR prior instead of the strongest competing prior | DenseNet BNN · CIFAR-10 adversarial examples (ϵ = 0.001) | validation accuracy (%): 78.9 -> 84.7 | Improved |
| Use the IR prior instead of the strongest competing prior | DenseNet BNN · CIFAR-100 adversarial examples (ϵ = 0.0001) | validation accuracy (%): 47.6 -> 50.2 | Improved |
Closest papers by meaning
- Deep Reference Priors: What is the best way to pretrain a model? (ICML 2022 · similarity 0.87)
- Amortising Inference and Meta-Learning Priors in Neural Networks (ICLR 2026 · similarity 0.87)
- Posterior Concentrations of Fully-Connected Bayesian Neural Networks with General Priors on the Weights (JMLR 2025 · similarity 0.87)
- Efficient and Scalable Bayesian Neural Nets with Rank-1 Factors (ICML 2020 · similarity 0.87)
- Make Me a BNN: A Simple Strategy for Estimating Bayesian Uncertainty from Pre-trained Models (CVPR 2024 · similarity 0.86)
- Training Bayesian Neural Networks with Sparse Subspace Variational Inference (ICLR 2024 · 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.
