Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Singular Bayesian Neural Networks
Mame Diarra Toure, David A. Stephens
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 |
|---|---|---|---|
| replace full-rank mean-field weights with low-rank factorization W=AB^T | BBB · MIMIC-III | AUC-OOD: 0.770 -> 0.802 | Improved |
| replace full-rank weights with low-rank factorization | BBB · MIMIC-III | parameters: 44.8k -> 13.6k | Improved |
| replace full-rank BBB with low-rank factorization | BBB · MIMIC-III | NLL: .401 -> .433 | Hurt |
| replace full-rank BBB with low-rank factorization | BBB · SST-2 | accuracy: 0.752 -> 0.806 | Improved |
| replace full-rank BBB with low-rank factorization | BBB · SST-2 | training time: 23.1 -> 8.2 minutes | Improved |
| replace full-rank with low-rank factorization | BBB · Beijing Air Quality | AUROC-OOD: 0.492 -> 0.710 | Improved |
| use SVD initialization instead of random initialization (ablation) | Low-Rank BBB · MIMIC-III | AUROC: 0.895 -> 0.898 | No clear effect |
| use SVD initialization instead of random initialization (ablation) | Low-Rank BBB · Beijing Air Quality | AUROC-OOD: 0.710 -> 0.687 | Hurt |
| use SVD initialization instead of random initialization (ablation) | Low-Rank BBB · MIMIC-III | AUC-OOD: .802 -> .713 | Hurt |
| use Low-Rank BBB instead of 5-member Deep Ensemble | Bayesian NN · MIMIC-III | AUC-OOD: 0.738 -> 0.802 | Improved |
| use Low-Rank BBB instead of 5-member Deep Ensemble | Bayesian NN · SST-2 | accuracy: 0.825 -> 0.806 | Hurt |
| use Low-Rank BBB instead of SWAG | Bayesian NN · SST-2 | NLL: 0.556 -> 0.527 | Improved |
Closest papers by meaning
- Scale-invariant Bayesian Neural Networks with Connectivity Tangent Kernel (ICLR 2023 · similarity 0.89)
- Deterministic PAC-Bayesian generalization bounds for deep networks via generalizing noise-resilience (ICLR 2019 · similarity 0.88)
- Efficient and Scalable Bayesian Neural Nets with Rank-1 Factors (ICML 2020 · similarity 0.87)
- Dissecting Non-Vacuous Generalization Bounds based on the Mean-Field Approximation (ICML 2020 · similarity 0.87)
- The k-tied Normal Distribution: A Compact Parameterization of Gaussian Mean Field Posteriors in Bayesian Neural Networks (ICML 2020 · similarity 0.87)
- A Symmetry-Aware Exploration of Bayesian Neural Network Posteriors (ICLR 2024 · 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.
