Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Bayesian Nested Neural Networks for Uncertainty Calibration and Adaptive Compression
Yufei Cui, Ziquan Liu, Qiao Li, Antoni B. Chan, Chun Jason Xue
Check 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 |
|---|---|---|---|
| Learn node importance and nested dropout rates from data rather than fixing them | BN3 versus FN3 · Cifar10 and Tiny ImageNet classification | accuracy | Improved |
| Use BN3 instead of independently trained Bayesian networks at different widths | BN3 and IBNN · Cifar10 and Tiny ImageNet classification | accuracy | Mixed |
| Use a Bayesian nested network rather than a deterministic nested network | BN3 versus FN3 · Cifar10 and Tiny ImageNet classification | ECE | Improved |
| Increase network width (ablation) | BN3 · Cifar10 and Tiny ImageNet classification | accuracy | Improved |
| Use the posterior mean instead of stochastic sampling at test time (ablation) | BN3 (DT-BN3 variant) · Cifar10 and Tiny ImageNet classification | accuracy | No clear effect |
| Average predictions from stochastic forward passes at different widths (ablation) | BN3 (BN3-MC variant) · Cifar10, VGG11 | accuracy | Hurt |
| Increase BN3 width from the smallest configuration (ablation) | BN3 · LIDC-IDRI lung segmentation | GED | Improved |
| Use the smallest BN3 width (ablation) | BN3 · LIDC-IDRI lung segmentation | GED | Hurt |
| Use temperature scaling on FN3 sub-networks (ablation) | FN3 · Cifar10 and Tiny ImageNet classification | ECE | Mixed |
| Use the smallest BN3 sub-network instead of SVB | BN3 and SVB · Cifar10, VGG11 | accuracy: 82.9 -> 83.5 | Improved |
| Use the smallest BN3 sub-network instead of SVB | BN3 and SVB · Cifar10, VGG11 | ECE: 0.17 -> 0.10 | Improved |
| Use the smallest BN3 sub-network instead of SVB | BN3 and SVB · Cifar10, VGG11 | OOD AUPR: 0.94 -> 0.94 | No clear effect |
Closest papers by meaning
- Variational Nested Dropout (TPAMI 2023 · similarity 0.90)
- Deep Hierarchical Learning with Nested Subspace Networks (ICLR 2026 · similarity 0.86)
- Training Bayesian Neural Networks with Sparse Subspace Variational Inference (ICLR 2024 · similarity 0.85)
- Bayesian Layers: A Module for Neural Network Uncertainty (NeurIPS 2019 · similarity 0.85)
- Make Me a BNN: A Simple Strategy for Estimating Bayesian Uncertainty from Pre-trained Models (CVPR 2024 · similarity 0.84)
- Parameter Expanded Stochastic Gradient Markov Chain Monte Carlo (ICLR 2025 · similarity 0.84)
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.
