Conditional normalizing flow for Markov chain Monte Carlo sampling in the critical region of lattice field theory
Ankur Singha, Dipankar Chakrabarti, Vipul Arora
Abstract
The traditional Markov chain Monte Carlo (MCMC) suffers from the problem of critical slowing down. Generative machine learning methods, such as normalizing flows, offer a promising method to speed up MCMC simulations, especially in critical regions of lattice field theory. However, training these models for different parameter values in the critical region is inefficient. In this paper, we address this issue by interpolating or extrapolating the flow model to any parameter value in the critical region. We demonstrate the effectiveness of the proposed method for MCMC sampling in critical regions.
Topics & Concepts
Markov chain Monte CarloStatistical physicsMetropolis–Hastings algorithmMonte Carlo methodHybrid Monte CarloMarkov chainMarkov chain mixing timeLattice (music)Computer scienceMonte Carlo method in statistical physicsPhysicsMathematicsMarkov modelMarkov propertyStatisticsMachine learningAcousticsMarkov Chains and Monte Carlo MethodsTheoretical and Computational PhysicsStochastic processes and statistical mechanics