Research map / Probabilistic and scientific modeling / Variational and generative sampling
Towards Unifying Hamiltonian Monte Carlo and 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 |
|---|---|---|---|
| increase monomial parameter a from 1 to 2 in MG-HMC (ablation) | MG-HMC on 1D bimodal distribution · 1D bimodal (U=x^4-2x^2) | ESS: 10157 -> 24298 | Helped |
| increase monomial parameter a from 1 to 2 in MG-HMC (ablation) | MG-HMC on 2D bimodal distribution · 2D bimodal | ESS: 16349 -> 18007 | Helped |
| set a=1 (standard HMC/slice-equivalent kinetic) rather than a=2 on real data (ablation) | MG-HMC on Bayesian logistic regression · BLR datasets minimum ESS | minimum ESS: a=1 best on A,G,H,P,R; a=2 worse on A (1490 vs 4308) and R and C | Mixed |
| set a=1 instead of a=0.5 or a=2 for ICA (ablation) | MG-HMC on MEG ICA · ICA (25 dim) minimum ESS | minimum ESS: 3029 vs 2677 (a=0.5) and 1534 (a=2) | Helped |
| increase monomial parameter a (heavier-tailed momentum) in analytic MG-SS on 1D unimodal targets | analytic MG-SS · Exponential distribution, one-step autocorrelation | rho_x(1): rho = 1/(a+1) decreases toward zero | Helped |
| increase monomial parameter a in MG-HMC with numerical integrator on 1D unimodal targets | MG-HMC (leapfrog) · Gamma distribution, one-step autocorrelation | rho_x(1): rho first decreases then increases with a | Mixed |
| increase monomial parameter a in MG-HMC with numerical integrator (acceptance rate effect) | MG-HMC on 1D unimodal targets · 1D unimodal (exponential, truncated Gaussian, Gamma) | acceptance rate: 0.98 -> 0.77 as a goes 0.5 to 4 (exponential/Gaussian) | Hurt |
| set a=1 instead of a=2 for high-dimensional targets (practical compromise) (ablation) | MG-HMC on Caravan (87 dim) BLR · BLR Caravan minimum ESS | minimum ESS: 36 (a=1) vs 7 (a=2) | Helped |
| apply step-size decay schedule for MG-HMC with large a | MG-HMC with a>1 · general convergence | qualitative burn-in/convergence | Helped |
Closest papers by meaning
- Slice Sampling on Hamiltonian Trajectories (ICML 2016 · similarity 0.92)
- Fixed-Distance Hamiltonian Monte Carlo (NeurIPS 2022 · similarity 0.90)
- Reflective Hamiltonian Monte Carlo: Mixing Analysis and Application to Sampling on Stiefel Manifold (ICML 2026 · similarity 0.89)
- Unbiased constrained sampling with Self-Concordant Barrier Hamiltonian Monte Carlo (NeurIPS 2023 · similarity 0.89)
- Metropolis Adjusted Microcanonical Hamiltonian Monte Carlo (NeurIPS 2025 · similarity 0.88)
- Diffusion-warm sampling of the XY model enables fast thermalization at scale (NeurIPS 2026 · similarity 0.87)
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.