Research map / Graph learning / Graph algorithms and clustering
Mappings for Marginal Probabilities with Applications to Models in Statistical Physics
Mehdi Molkaraie
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 |
|---|---|---|---|
| run Gibbs sampling in the dual NFG instead of the primal NFG (large s=40) | Gaussian MRF marginal variance · 15x15 GMRF, s=40 | relative error of marginal variance: relative error of about 1.5\times 10^{-3} with only 10^2 samples | Improved |
| run Gibbs sampling in the dual NFG instead of the primal NFG (small s=1) | Gaussian MRF marginal variance · 15x15 GMRF, s=1 | convergence | Hurt |
| increase intervariable variance s from 1 to 20 to 40 (ablation) | Gibbs sampling on GMRF · 15x15 GMRF | convergence: exact variance 0.5589 -> 20.2046 -> 23.5498 | Mixed |
| run BP in the dual NFG instead of the primal NFG for strong couplings | BP on ferromagnetic Ising model · 2D Ising 6x6 zero field, sigma^2 > 0.25 | relative error of edge marginal: more than two orders of magnitude lower relative error | Improved |
| run BP in the dual NFG instead of primal NFG for large couplings | BP on fully connected Ising model · fully connected Ising N=10, betaJ_x > 0.20 | relative error of edge marginal: even by more than two orders of magnitude | Improved |
| run BP in dual domain on frustrated 3-state Potts model | BP on Potts model · 2D 3-state Potts 6x6, betaJ_Ferr > 1.90 | relative error of edge marginal | Improved |
| use TEP in primal domain instead of BP in dual | Potts model inference · 2D 3-state Potts 6x6, betaJ_Ferr < 1.0 | relative error of edge marginal | Improved |
| use TEP instead of BP in the dual domain | Ising model inference · 2D Ising 6x6 zero field | relative error of edge marginal | No clear effect |
Closest papers by meaning
- Sparse Estimation in Ising Model via Penalized Monte Carlo Methods (JMLR 2018 · similarity 0.86)
- Fast Wasserstein rates for estimating probability distributions of probabilistic graphical models (NeurIPS 2026 · similarity 0.85)
- Multi-Marginal Couplings for Metropolis--Hastings (NeurIPS 2026 · similarity 0.84)
- Generalized Direct Change Estimation in Ising Model Structure (ICML 2016 · similarity 0.84)
- Learning Non-Gaussian Graphical Models via Hessian Scores and Triangular Transport (JMLR 2024 · similarity 0.84)
- A Machine Learning Approach to Duality in Statistical Physics (ICML 2025 · similarity 0.84)
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.
