Population annealing: Massively parallel simulations in statistical physics

Martin Weigel, Lev Yu Barash, Michal Borovský, Wolfhard Janke, Lev N. Shchur

Research output: Contribution to journalArticle

Abstract

The canonical technique for Monte Carlo simulations in statistical physics is importance sampling via a suitably constructed Markov chain. While such approaches are quite successful, they are not particularly well suited for parallelization as the chain dynamics is sequential, and if replicated chains are used to increase statistics each of them relaxes into equilibrium with an intrinsic time constant that cannot be reduced by parallel work. Population annealing is a sequential Monte Carlo method that simulates an ensemble of system replica under a cooling protocol. The population element makes it naturally well suited for massively parallel simulations, and bias can be systematically reduced by increasing the population size. We present an implementation of population annealing on graphics processing units and discuss its behavior for different systems undergoing continuous and first-order phase transitions.
Original languageEnglish
JournalJournal of Physics: Conference Series
Volume921
DOIs
Publication statusPublished - 15 Dec 2017

Fingerprint

physics
annealing
simulation
Markov chains
replicas
time constant
Monte Carlo method
sampling
statistics
cooling

Cite this

Population annealing: Massively parallel simulations in statistical physics. / Weigel, Martin; Barash, Lev Yu; Borovský, Michal; Janke, Wolfhard; Shchur, Lev N.

In: Journal of Physics: Conference Series, Vol. 921, 15.12.2017.

Research output: Contribution to journalArticle

Weigel, Martin ; Barash, Lev Yu ; Borovský, Michal ; Janke, Wolfhard ; Shchur, Lev N. / Population annealing: Massively parallel simulations in statistical physics. In: Journal of Physics: Conference Series. 2017 ; Vol. 921.
@article{e9e0725c3f2247f9b183453daece7b3b,
title = "Population annealing: Massively parallel simulations in statistical physics",
abstract = "The canonical technique for Monte Carlo simulations in statistical physics is importance sampling via a suitably constructed Markov chain. While such approaches are quite successful, they are not particularly well suited for parallelization as the chain dynamics is sequential, and if replicated chains are used to increase statistics each of them relaxes into equilibrium with an intrinsic time constant that cannot be reduced by parallel work. Population annealing is a sequential Monte Carlo method that simulates an ensemble of system replica under a cooling protocol. The population element makes it naturally well suited for massively parallel simulations, and bias can be systematically reduced by increasing the population size. We present an implementation of population annealing on graphics processing units and discuss its behavior for different systems undergoing continuous and first-order phase transitions.",
author = "Martin Weigel and Barash, {Lev Yu} and Michal Borovsk{\'y} and Wolfhard Janke and Shchur, {Lev N.}",
year = "2017",
month = "12",
day = "15",
doi = "10.1088/1742-6596/921/1/012017",
language = "English",
volume = "921",
journal = "Journal of Physics: Conference Series",
issn = "1742-6588",
publisher = "IOP Publishing",

}

TY - JOUR

T1 - Population annealing: Massively parallel simulations in statistical physics

AU - Weigel, Martin

AU - Barash, Lev Yu

AU - Borovský, Michal

AU - Janke, Wolfhard

AU - Shchur, Lev N.

PY - 2017/12/15

Y1 - 2017/12/15

N2 - The canonical technique for Monte Carlo simulations in statistical physics is importance sampling via a suitably constructed Markov chain. While such approaches are quite successful, they are not particularly well suited for parallelization as the chain dynamics is sequential, and if replicated chains are used to increase statistics each of them relaxes into equilibrium with an intrinsic time constant that cannot be reduced by parallel work. Population annealing is a sequential Monte Carlo method that simulates an ensemble of system replica under a cooling protocol. The population element makes it naturally well suited for massively parallel simulations, and bias can be systematically reduced by increasing the population size. We present an implementation of population annealing on graphics processing units and discuss its behavior for different systems undergoing continuous and first-order phase transitions.

AB - The canonical technique for Monte Carlo simulations in statistical physics is importance sampling via a suitably constructed Markov chain. While such approaches are quite successful, they are not particularly well suited for parallelization as the chain dynamics is sequential, and if replicated chains are used to increase statistics each of them relaxes into equilibrium with an intrinsic time constant that cannot be reduced by parallel work. Population annealing is a sequential Monte Carlo method that simulates an ensemble of system replica under a cooling protocol. The population element makes it naturally well suited for massively parallel simulations, and bias can be systematically reduced by increasing the population size. We present an implementation of population annealing on graphics processing units and discuss its behavior for different systems undergoing continuous and first-order phase transitions.

U2 - 10.1088/1742-6596/921/1/012017

DO - 10.1088/1742-6596/921/1/012017

M3 - Article

VL - 921

JO - Journal of Physics: Conference Series

JF - Journal of Physics: Conference Series

SN - 1742-6588

ER -