Taming correlations through entropy-efficient measure decompositions with applications to mean-field approximation

DOI10.1007/S00440-019-00924-2MaRDI QIDQ2174660zbMATH OpenOpenAlexWikidataFDO

Authors Ronen Eldan

Publication date 21 April 2020

Published in Zeitschrift für Wahrscheinlichkeitstheorie und Verwandte Gebiete (Search for Journal in Brave)

Full work available at URL https://arxiv.org/abs/1811.11530

Mathematics Subject Classification ID

Geometric probability and stochastic geometry (60D05) Applications of stochastic analysis (to PDEs, etc.) (60H30) Interacting random processes; statistical mechanics type models; percolation theory (60K35) Dynamic lattice systems (kinetic Ising, etc.) and systems on graphs in time-dependent statistical mechanics (82C20) Interacting particle systems in time-dependent statistical mechanics (82C22)

Abstract: The analysis of various models in statistical physics relies on the existence of decompositions of measures into mixtures of product-like components, where the goal is to attain a decomposition into measures whose entropy is close to that of the original measure, yet with small correlations between coordinates. We prove a related general result: For every isotropic measure $mu$ on $mathbb{R}^n$ and every $epsilon>0$, there exists a decomposition $mu=intm{u}_{h}etadm(heta)$ such that $H(mu)-mathbb{E}_{hetasimm}H(m{u}_{h}eta)leqnepsilon$ and $mathbb{E}_{hetasimm}mathrmCov(m{u}_{h}eta)preceqmathrmId/epsilon$. As an application, we prove a general bound for the mean-field approximation of Ising and Potts models, which is in a sense dimension free, in both continuous and discrete settings. In particular, for an Ising model on ${pm1}^n$ or on $[-1,1{]}^n$, we show that the deficit between the mean-field approximation and the free energy is at most $Cfrac1+ppleft(n|J{|}_{S_p}ight{)}^{fracp1+p}$ for all $p>0$, where $|J{|}_{S_p}$ denotes the Schatten-$p$ norm of the interaction matrix. For the case $p=2$, this recovers the result of [Jain et al., 2018], but for an optimal choice of $p$ it often allows to get almost dimension-free bounds.

Recommendations

• $\Phi$ -Entropic Measures of Correlation
• Mean-field approximation, convex hierarchies, and the optimality of correlation rounding: a unified perspective
• On the entropic measures of stochastic dependence
• A maximum likelihood approach to correlation dimension and entropy estimation
• Entropic extensivity and large deviations in the presence of strong correlations
• Approximating pairwise correlations in the Ising model
• Decomposition of mean-field Gibbs distributions into product measures
• Exact decay of correlations for mixtures and rotators
• Fluctuation and entropy in spectrally constrained random fields
• On the decay of correlations in the random field Ising model

Cites work

• A birthday repetition theorem and complexity of approximating dense CSPs
• A transportation approach to the mean-field approximation
• Approximating CSPs with global cardinality constraints using SDP hierarchies
• Bethe states of random factor graphs
• Decomposition of mean-field Gibbs distributions into product measures
• Factor models on locally tree-like graphs
• Gaussian-width gradient complexity, reverse log-Sobolev inequalities and nonlinear large deviations
• Harnessing the Bethe free energy
• Mean-field approximation, convex hierarchies, and the optimality of correlation rounding: a unified perspective
• Nonlinear large deviations
• Skorokhod embeddings via stochastic flows on the space of Gaussian measures
• The CLT in high dimensions: quantitative bounds via martingale embedding
• The structure of low-complexity Gibbs measures on product spaces
• Thin shell implies spectral gap up to polylog via a stochastic localization scheme
• Universality of the mean-field for the Potts model

Cited in

(14)

• Decomposition of mean-field Gibbs distributions into product measures
• A transportation approach to the mean-field approximation
• Log concavity and concentration of Lipschitz functions on the Boolean hypercube
• The cut metric for probability distributions
• Mean-field approximation, convex hierarchies, and the optimality of correlation rounding: a unified perspective
• Fluctuations in mean-field Ising models
• Analysis of high-dimensional distributions using pathwise methods
• Mean field approximations via log-concavity
• Causal effect estimation under network interference with mean-field methods
• Expanderizing higher-order random walks
• Expanderizing higher-order random walks
• Sampling from mean-field Gibbs measures via diffusion processes
• Localization schemes: a framework for proving mixing bounds for Markov chains
• Rapid Bayesian computation and estimation for neural networks via log-concave coupling

This page was built for publication: Taming correlations through entropy-efficient measure decompositions with applications to mean-field approximation

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2174660)