Research map / Probabilistic and scientific modeling / Variational and generative sampling
Minibatch Gibbs Sampling on Large Graphical Models
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 |
|---|---|---|---|
| Estimate total energy with a bias-adjusted Poisson minibatch and cache the current estimate | MIN-Gibbs · Factor graphs | compute cost/iteration: $O(D \Delta) -> O(D \Psi^2)$ | Helped |
| Use a minibatched Gibbs proposal with a Metropolis-Hastings correction | MGPMH · Factor graphs | compute cost/iteration: $O(D \Delta) -> O(D L^2 + \Delta)$ | Helped |
| Combine minibatched proposals with a second minibatch energy estimate | DoubleMIN-Gibbs · Factor graphs | compute cost/iteration: $O(D \Delta) -> O(D L^2 + \Psi^2)$ | Helped |
| Increase the MIN-Gibbs minibatch size | MIN-Gibbs · Fully connected Ising model | marginal-estimate convergence | Helped |
| Increase the local minibatch size | Local Minibatch Gibbs · Fully connected Ising model | marginal-estimate convergence | Helped |
| Increase the MGPMH minibatch size | MGPMH · Fully connected Potts model | marginal-estimate convergence | Helped |
| Increase the second minibatch size | DoubleMIN-Gibbs · Fully connected Potts model | marginal-estimate convergence | Helped |
| Use an energy estimator whose error is bounded by $\delta$ | MIN-Gibbs · Factor graphs | spectral gap relative to Gibbs: $\bar \gamma \ge \exp(-6 \delta) \cdot \gamma$ | Hurt |
| Use a minibatch size satisfying $L \le \lambda$ | MGPMH · Factor graphs | spectral gap relative to Gibbs: $\bar \gamma \ge \exp\left(-L^2 / \lambda\right) \cdot \gamma$ | Hurt |
| Use a second energy estimator whose error is bounded by $\delta$ | DoubleMIN-Gibbs · Factor graphs | spectral gap relative to MGPMH: $\bar \gamma \ge \exp(-4 \delta) \cdot \gamma$ | Hurt |
Closest papers by meaning
- Poisson-Minibatching for Gibbs Sampling with Convergence Rate Guarantees (NeurIPS 2019 · similarity 0.95)
- Closed-Form Gibbs Sampling for Graphical Models with Algebraic Constraints (AAAI 2016 · similarity 0.89)
- Diffusive Gibbs Sampling (ICML 2024 · similarity 0.86)
- Scalable Discrete Diffusion Samplers: Combinatorial Optimization and Statistical Physics (ICLR 2025 · similarity 0.86)
- Composing Diffusion Priors with Explicit Physical Context via Generative Gibbs Sampling (NeurIPS 2026 · similarity 0.86)
- Multi-Marginal Couplings for Metropolis--Hastings (NeurIPS 2026 · similarity 0.86)
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.