Scalable generative sampling and multilevel estimation for lattice field theories near criticality
6 affiliations
- 1Berlin Institute for the Foundations of Learning and Data (BIFOLD)
- 2Technische Universität Berlin
- 3The University of Edinburgh
- 4Deutsches Elektronen-Synchrotron (DESY)
- 5The Cyprus Institute
- 6RIKEN Center for Advanced Intelligence Project (AIP)
Abstract
Sampling lattice field theories near criticality is severely hindered by critical slowing down, which makes standard Markov chain methods increasingly inefficient at large lattice volumes. We introduce a multiscale generative sampler, inspired by renormalization-group ideas, that models the Boltzmann distribution through a coarse-to-fine hierarchy across length scales. At each level, a conditional Gaussian mixture model captures the main local dependence of newly introduced variables on the already-sampled coarse field, while a masked continuous normalizing flow refines the remaining conditional structure. Coarse levels encode the dominant long-wavelength modes, and finer levels progressively add short-distance fluctuations. In addition, because the architecture preserves coarse fields exactly during refinement, it provides exact restriction maps at no additional computational cost and directly enables asymptotically unbiased multilevel Monte Carlo variance reduction. For the two-dimensional scalar ϕ 4 theory near criticality, the method achieves integrated autocorrelation times orders of magnitude smaller than hybrid Monte Carlo (HMC) on large volumes, maintains high importance-sampling efficiency relative to other generative baselines, and reproduces asymptotically unbiased physical observables in statistical agreement with long HMC simulations.
Related
- Sampling from the Sherrington-Kirkpatrick Model up to Criticality
- Multilevel Plaquette-Space Sampling for Lattice Gauge Theories with Local Constraint Solves
- Scalable generative modeling of non-Gaussian spatio-temporal fields via autoregressive Gaussian processes
- Computationally efficient interval estimation for the 1-1-1 multilevel mediation analysis model in R: A replication of Bauer, Preacher, and Gil (2006)
- Scalable Logistic Gaussian Process Density Regression with Kinetic Langevin Sampling
BibTeX
@article{singha2026scalable,
title = {{Scalable generative sampling and multilevel estimation for lattice field theories near criticality}},
author = {A. Singha and J. Kauffmann and E. Cellini and K. Jansen and S. Nakajima},
journal = {Physical review. D/Physical review. D.},
year = {2026},
month = oct,
doi = {10.1103/jjss-lwzv},
url = {https://doi.org/10.1103/jjss-lwzv},
note = {CC BY},
}
RIS
TY - JOUR
TI - Scalable generative sampling and multilevel estimation for lattice field theories near criticality
AU - A. Singha
AU - J. Kauffmann
AU - E. Cellini
AU - K. Jansen
AU - S. Nakajima
PY - 2026
DA - 2026/10/05
JO - Physical review. D/Physical review. D.
DO - 10.1103/jjss-lwzv
UR - https://doi.org/10.1103/jjss-lwzv
AB - Sampling lattice field theories near criticality is severely hindered by critical slowing down, which makes standard Markov chain methods increasingly inefficient at large lattice volumes. We introduce a multiscale generative sampler, inspired by renormalization-group ideas, that models the Boltzmann distribution through a coarse-to-fine hierarchy across length scales. At each level, a conditional Gaussian mixture model captures the main local dependence of newly introduced variables on the already-sampled coarse field, while a masked continuous normalizing flow refines the remaining conditional structure. Coarse levels encode the dominant long-wavelength modes, and finer levels progressively add short-distance fluctuations. In addition, because the architecture preserves coarse fields exactly during refinement, it provides exact restriction maps at no additional computational cost and directly enables asymptotically unbiased multilevel Monte Carlo variance reduction. For the two-dimensional scalar ϕ 4 theory near criticality, the method achieves integrated autocorrelation times orders of magnitude smaller than hybrid Monte Carlo (HMC) on large volumes, maintains high importance-sampling efficiency relative to other generative baselines, and reproduces asymptotically unbiased physical observables in statistical agreement with long HMC simulations.
SN - 2470-0010
SN - 2470-0029
SN - 2470-0037
ER -