Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Nonlinear MCMC for Bayesian Machine Learning
James Vuckovic
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 linear MCMC with nonlinear BG | Bayesian ResNet-18 · CIFAR10 | ECE (×10^2), non-tempered: 0.24_{\pm0.02} -> 0.14_{\pm0.03} | Improved |
| Replace linear MCMC with nonlinear BG | Bayesian ResNet-18 · CIFAR10 | ECE (×10^2), tempered: 0.26_{\pm0.014} -> 0.16_{\pm0.03} | Improved |
| Replace linear MCMC with nonlinear BG | Bayesian ResNet-18 · CIFAR10 | MCE, non-tempered: 3.78_{\pm0.43} -> 3.21_{\pm0.75} | No clear effect |
| Replace linear MCMC with nonlinear BG | Bayesian ResNet-18 · CIFAR10 | MCE, tempered: 4.88_{\pm0.85} -> 3.72_{\pm0.40} | Improved |
| Add tempering to nonlinear AR (ablation) | Bayesian ResNet-18 · CIFAR10 | ECE: Diverged -> 0.15_{\pm0.05} | Improved |
| Add tempering to nonlinear AR (ablation) | Bayesian ResNet-18 · CIFAR10 | MCE: Diverged -> 3.45_{\pm 1.09} | Improved |
| Use nonlinear AR instead of linear MCMC | Bayesian ResNet-18 · CIFAR10 | ECE, non-tempered: Diverged | Hurt |
| Use nonlinear MCMC instead of linear MCMC | Toy sampling algorithms · Circular MoG | Wall time (s): 0.97_{\pm 0.02} -> 2.57_{\pm 0.06} (AR); 2.99_{\pm 0.01} (BG) | Hurt |
| Use nonlinear MCMC instead of linear MCMC | Toy sampling algorithms · Two Rings | Wall time (s): 0.97_{\pm 0.02} -> 2.57_{\pm 0.06} (AR); 3.18_{\pm 0.01} (BG) | Hurt |
| Use nonlinear MCMC instead of linear MCMC | Toy sampling algorithms · Grid MoG | Wall time (s): 1.33_{\pm 0.01} -> 2.66_{\pm 0.01} (AR); 3.70_{\pm 0.01} (BG) | Hurt |
| Increase nonlinear sampler particle count (ablation) | Nonlinear MCMC · Circular MoG, Two Rings, and Grid MoG | MMD-squared | Improved |
Closest papers by meaning
- Parameter Expanded Stochastic Gradient Markov Chain Monte Carlo (ICLR 2025 · similarity 0.87)
- Meta-Learning For Stochastic Gradient MCMC (ICLR 2019 · similarity 0.87)
- Structured Stochastic Gradient MCMC (ICML 2022 · similarity 0.87)
- Microcanonical Langevin Ensembles: Advancing the Sampling of Bayesian Neural Networks (ICLR 2025 · similarity 0.85)
- Learning to Explore for Stochastic Gradient MCMC (ICML 2024 · similarity 0.85)
- Parallelizing MCMC Across the Sequence Length (NeurIPS 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.
