Optimized population Monte Carlo
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866917568127696896 |
|---|---|
| author | Ebert, P. L. Gessert, D. Janke, W. Weigel, M. |
| author_facet | Ebert, P. L. Gessert, D. Janke, W. Weigel, M. |
| contents | Population Monte Carlo simulations in the form commonly referred to as population annealing can serve as a useful meta-algorithm for simulating systems with complex free-energy landscapes. In the present paper we provide an easily accessible introduction to the approach, focusing on spin systems as simple example problems. While the method is very general and powerful, it also comes with a number of tunable parameters. Here, we discuss the question of an optimal choice of resampling protocol, that is shown to have significant influence on the quality of results. While population annealing is an excellent fit to the paradigm of massively parallel simulations, limitations in the availability of parallel resources and especially memory can provide a bottleneck to its efficacy. As we demonstrate for results of the Ising ferromagnetic and spin-glass models, weighted averages of smaller-scale runs can be easily combined to reduce both systematic and statistical errors in order to avoid such bottlenecks. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2401_07965 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Optimized population Monte Carlo Ebert, P. L. Gessert, D. Janke, W. Weigel, M. Statistical Mechanics Disordered Systems and Neural Networks Computational Physics Population Monte Carlo simulations in the form commonly referred to as population annealing can serve as a useful meta-algorithm for simulating systems with complex free-energy landscapes. In the present paper we provide an easily accessible introduction to the approach, focusing on spin systems as simple example problems. While the method is very general and powerful, it also comes with a number of tunable parameters. Here, we discuss the question of an optimal choice of resampling protocol, that is shown to have significant influence on the quality of results. While population annealing is an excellent fit to the paradigm of massively parallel simulations, limitations in the availability of parallel resources and especially memory can provide a bottleneck to its efficacy. As we demonstrate for results of the Ising ferromagnetic and spin-glass models, weighted averages of smaller-scale runs can be easily combined to reduce both systematic and statistical errors in order to avoid such bottlenecks. |
| title | Optimized population Monte Carlo |
| topic | Statistical Mechanics Disordered Systems and Neural Networks Computational Physics |
| url | https://arxiv.org/abs/2401.07965 |