Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Efficient and Scalable Bayesian Neural Nets with Rank-1 Factors
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 rank-1 BNNs instead of BatchEnsemble | Wide ResNet 28-10 · CIFAR-10-C | NLL: 1.02 to 0.74 | Helped |
| Use rank-1 BNNs instead of BatchEnsemble | Wide ResNet 28-10 · CIFAR-10-C | accuracy: +3.7% | Helped |
| Use rank-1 BNNs instead of BatchEnsemble | Wide ResNet 28-10 · CIFAR-10-C | ECE: decreases by 0.05 | Helped |
| Use a Cauchy rather than Gaussian rank-1 prior and posterior | Rank-1 BNN · CIFAR-100 | accuracy | Helped |
| Use a Gaussian rather than Cauchy rank-1 prior and posterior | Rank-1 Bayesian RNN · MIMIC-III | AUC-PR | Helped |
| Increase the mixture size from 1 to 8 components (ablation) | Rank-1 BNN · CIFAR-10 and CIFAR-10-C | NLL | Helped |
| Increase the mixture size to 16 components (ablation) | Rank-1 BNN · CIFAR-10 and CIFAR-10-C | NLL | Hurt |
| Train with log-mixture likelihood rather than average per-component log-likelihood (ablation) | Rank-1 BNN · CIFAR-10 | NLL | Hurt |
| Place priors on both rank-1 factor vectors rather than only one (ablation) | Rank-1 BNN · CIFAR-10-C | NLL | Helped |
| Initialize variational standard deviations close to zero (ablation) | Rank-1 BNN · test set | accuracy | Helped |
| Initialize variational standard deviations close to zero (ablation) | Rank-1 BNN · test set | ECE | Hurt |
| Use more Monte Carlo samples during evaluation (ablation) | Gaussian Rank-1 BNN · image classification tasks | performance | Helped |
Closest papers by meaning
- Bayesian Low-Rank Learning (Bella): A Practical Approach to Bayesian Neural Networks (AAAI 2025 · similarity 0.91)
- What Are Bayesian Neural Network Posteriors Really Like? (ICML 2021 · similarity 0.90)
- A Symmetry-Aware Exploration of Bayesian Neural Network Posteriors (ICLR 2024 · similarity 0.90)
- Make Me a BNN: A Simple Strategy for Estimating Bayesian Uncertainty from Pre-trained Models (CVPR 2024 · similarity 0.89)
- Training Bayesian Neural Networks with Sparse Subspace Variational Inference (ICLR 2024 · similarity 0.89)
- A Study of Bayesian Neural Network Surrogates for Bayesian Optimization (ICLR 2024 · similarity 0.88)
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.