Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Richer Bayesian Last Layers with Subsampled NTK Features
Sergio Calvo Ordoñez, Jonathan Plenk, Richard Bergna, Alvaro Cartea, Yarin Gal, José Miguel Hernández-Lobato, Kamil Ciosek
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 |
|---|---|---|---|
| add NTK low-rank correction (Rich-BLL) to the Bayesian last layer | NNGP (BLL) · UCI Concrete | test NLL: 3.39 -> 3.10 | Improved |
| add NTK low-rank correction (Rich-BLL) to the Bayesian last layer | NNGP (BLL) · UCI Boston | test NLL: 2.78 -> 2.61 | Improved |
| add NTK low-rank correction (Rich-BLL) to the Bayesian last layer | NNGP (BLL) · UCI Energy | test NLL: 0.92 -> 0.75 | Improved |
| add subsampling (40% of data) to Rich-BLL (ablation) | Rich-BLL · UCI regression | test NLL: 0.75 -> 0.78 | No clear effect |
| add NTK correction to the Bayesian last layer | NNGP (BLL) · Wheel bandit delta=0.99 | normalized cumulative regret: 55.80 -> 21.80 | Improved |
| add subsampling to Rich-BLL (ablation) | Rich-BLL · Wheel bandit delta=0.99 | normalized cumulative regret: 21.80 -> 22.30 | No clear effect |
| add NTK correction to the Bayesian last layer | NNGP (BLL) · Wine Quality OOD | AUROC: 0.88 -> 0.96 | Improved |
| add NTK correction (subsampled) to the Bayesian last layer | NNGP (BLL) · CIFAR-10 | NLL: 0.58 -> 0.56 | Improved |
| add NTK correction (subsampled) to the Bayesian last layer | NNGP (BLL) · CIFAR-10 vs SVHN | AUROC: 0.88 -> 0.91 | Improved |
| add NTK correction (subsampled) to the Bayesian last layer | NNGP (BLL) · CIFAR-10 vs CIFAR-100 | AUROC: 0.62 -> 0.65 | Improved |
| use a single global transformation instead of class-specific ones (ablation) | Rich-BLL classification · CIFAR-10 | accuracy/uncertainty | Improved |
Closest papers by meaning
- Variational Bayesian Last Layers (ICLR 2024 · similarity 0.90)
- Uncertainty Quantification with the Empirical Neural Tangent Kernel (NeurIPS 2025 · similarity 0.87)
- Make Me a BNN: A Simple Strategy for Estimating Bayesian Uncertainty from Pre-trained Models (CVPR 2024 · similarity 0.86)
- Transitional Uncertainty with Layered Intermediate Predictions (ICML 2024 · similarity 0.86)
- A Symmetry-Aware Exploration of Bayesian Neural Network Posteriors (ICLR 2024 · similarity 0.86)
- Enhancing Uncertainty Estimation and Interpretability with Bayesian Non-negative Decision Layer (ICLR 2025 · 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.
