Research map / Probabilistic and scientific modeling / Variational and generative sampling
Gibbsian Polar Slice Sampling
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 |
|---|---|---|---|
| Gibbsian polar slice sampling (GPSS) with stepping-out and shrinkage vs. ESS and HRUSS | multivariate standard Cauchy d=100 · Multivariate Cauchy d=100 | convergence of p(b) estimates over iterations | Helped |
| Gibbsian polar slice sampling vs. tuned/untuned ESS and HRUSS | hyperplane disk d=200 · Hyperplane Disk d=200 | mean step size: GPSS ≈5.0 vs tuned ESS ≈3.5, untuned ESS ≈2.4, HRUSS ≈0.6 | Helped |
| Gibbsian polar slice sampling vs. ESS and HRUSS | axial modes, d=10 to 100 · Axial modes | mean dwelling time | Helped |
| Gibbsian polar slice sampling vs. tuned ESS on Neal's funnel | Neal's funnel d=10 with 1-minute time budget · Neal's funnel d=10 | convergence of mean, std, 0.001 and 0.999 quantiles over runtime | Helped |
| Gibbsian polar slice sampling vs. NUTS on Neal's funnel | Neal's funnel d=10 · Neal's funnel d=10 | 0.001-quantile, mean and std estimation | Helped |
| Radius update with dedicated direction update (GPSS) vs. HRUSS on hyperplane disk | hyperplane disk d=200 · Hyperplane Disk d=200 | integrated autocorrelation time of radii: GPSS 1.09 vs HRUSS 684.57 | Helped |
| Tuning ESS covariance from GPSS samples (tuned ESS) vs. untuned ESS in hyperplane experiment (ablation) | ESS on hyperplane disk d=200 · Hyperplane Disk d=200 | integrated autocorrelation time of radii: tuned 188.13 vs naive 199.17 | Helped |
| Tuning ESS covariance (sophisticated vs naive) in axial modes experiment (ablation) | ESS on axial modes · Axial modes | mean dwelling time and maximum dwelling time | Helped |
| Stepping-out and shrinkage GPSS on heavy-tailed Cauchy: GPSS is cheaper per iteration than HRUSS/ESS in target evaluations | GPSS · Multivariate Cauchy d=100 | target density evaluations per iteration: GPSS 6.90 vs HRUSS 8.46, naive ESS 5.86 | No clear effect |
| Neal's funnel iterations per time budget: GPSS completes fewer iterations than HRUSS and about the same as ESS | GPSS · Neal's funnel d=10 | iterations completed in 1 minute: 3.39e5 vs HRUSS 7.95e5 | Hurt |
| Covariance-tuned ESS vs GPSS in hyperplane experiment: ESS slower per iteration | tuned ESS · Hyperplane Disk d=200 | target density evaluations per iteration and runtime: TDE/I 6.51 vs GPSS 12.23 (GPSS uses more evals, but tuned ESS consistently ran slower overall) | Hurt |
Closest papers by meaning
- Diffusive Gibbs Sampling (ICML 2024 · similarity 0.86)
- Parallel Affine Transformation Tuning of Markov Chain Monte Carlo (ICML 2024 · similarity 0.86)
- Chain of Log-Concave Markov Chains (ICLR 2024 · similarity 0.86)
- Split Gibbs Discrete Diffusion Posterior Sampling (NeurIPS 2025 · similarity 0.86)
- Faster Sampling from Log-Concave Densities over Polytopes via Efficient Linear Solvers (ICLR 2024 · similarity 0.86)
- Restricted Spectral Gap Decomposition for Simulated Tempering Targeting Mixture Distributions (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.