Conditional normalizing flow for Markov chain Monte Carlo sampling in the critical region of lattice field theory

Singha, Ankur; Chakrabarti, Dipankar; Arora, Vipul

doi:10.1103/physrevd.107.014512

Cited by 20 publications

(13 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A simple scheme to condition a split-coupling flow architecture was proposed by Ardizzone et al [39] in relation to image generation. These ideas were then applied to physical problems in the context of lattice field theory by Gerdes et al [31] and Singha et al [32,33]. In the model proposed by Ardizzone et al [39], the transformation is conditioned by concatenating the condition data c to the coupling layer network input, as indicated in figure 2.…”

Section: Targeted Sampling Using Boltzmann Generatorsmentioning

confidence: 99%

“…These models are of particular interest since they allow for unbiased estimation of physical observables [30]. More recently, conditioned flow-based models have been successfully applied in the context of physical problems to enhance sampling in systems studied using lattice field theory [31][32][33]. In this work, we propose a parallel sampling scheme to explore the reactive path space based on normalizing flows in the form of Boltzmann generators [22] for the generation of shooting points.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Conditioning Boltzmann generators for rare event sampling

Falkner,

Coretti,

Romano

et al. 2023

Mach. Learn.: Sci. Technol.

View full text Add to dashboard Cite

Understanding the dynamics of complex molecular processes is often linked to the study of infrequent transitions between long-lived stable states. The standard approach to the sampling of such rare events is to generate an ensemble of transition paths using a random walk in trajectory space. This, however, comes with the drawback of strong correlations between subsequently sampled paths and with an intrinsic difficulty in parallelizing the sampling process. We propose a transition path sampling scheme based on neural-network generated configurations. These are obtained employing normalizing flows, a neural network class able to generate statistically independent samples from a given distribution. With this approach, not only are correlations between visited paths removed, but the sampling process becomes easily parallelizable. Moreover, by conditioning the normalizing flow, the sampling of configurations can be steered towards regions of interest. We show that this approach enables the resolution of both the thermodynamics and kinetics of the transition region for systems that can be sampled using exact-likelihood generative models.

show abstract

Section: Targeted Sampling Using Boltzmann Generatorsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Conditioning Boltzmann generators for rare event sampling

Falkner,

Coretti,

Romano

et al. 2023

Mach. Learn.: Sci. Technol.

View full text Add to dashboard Cite

show abstract

“…By now, many works have demonstrated flow-based sampling for scalar field theory with larger lattice sizes, smaller lattice spacings, and in the broken symmetry phase [32][33][34][35][36][37][38][39][40]. Since this early work, there have also been many developments in the direction of incorporating gauge symmetries, incorporating dynamical fermions, and scaling to address challenging problems in lattice QCD, as we now discuss.…”

Section: Application To Scalar Field Theorymentioning

confidence: 99%

Flow-based sampling for fermionic lattice field theories

et al. 2021

View full text Add to dashboard Cite

Critical slowing down and topological freezing severely hinder Monte Carlo sampling of lattice field theories as the continuum limit is approached. Recently, significant progress has been made in applying a class of generative machine learning models, known as "flow-based" samplers, to combat these issues. These generative samplers also enable promising practical improvements in Monte Carlo sampling, such as fully parallelized configuration generation. These proceedings review the progress towards this goal and future prospects of the method.

show abstract

“…Using normalizing flows for sampling in lattice field theory has gained significant attention over the last few years. Several works have been carried out in the domain of scalar field theories [1][2][3][4][5][6][7][8][9][10], U (1) [11,12] and SU (N ) [13][14][15] pure gauge theories, and fermionic gauge theories [16,17]. This rapid development is attributed to the appealing conceptual properties of flow-based sampling.…”

Section: Introductionmentioning

confidence: 99%