A reinforcement learning strategy to automate and accelerate h/p-multigrid solvers
We explore a reinforcement learning strategy to automate and accelerate h/p-multigrid methods in high-order solvers. Multigrid methods are very efficient but require fine-tuning of numerical parameters, such as the number of smoothing sweeps per level and the correction fraction (i.e., proportion of...
Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Elsevier
2024-12-01
|
| Series: | Results in Engineering |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S2590123024012040 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1846115910841008128 |
|---|---|
| author | David Huergo Laura Alonso Saumitra Joshi Adrian Juanicotena Gonzalo Rubio Esteban Ferrer |
| author_facet | David Huergo Laura Alonso Saumitra Joshi Adrian Juanicotena Gonzalo Rubio Esteban Ferrer |
| author_sort | David Huergo |
| collection | DOAJ |
| description | We explore a reinforcement learning strategy to automate and accelerate h/p-multigrid methods in high-order solvers. Multigrid methods are very efficient but require fine-tuning of numerical parameters, such as the number of smoothing sweeps per level and the correction fraction (i.e., proportion of the corrected solution that is transferred from a coarser grid to a finer grid). The objective of this paper is to use a proximal policy optimization algorithm to automatically tune the multigrid parameters and, by doing so, improve stability and efficiency of the h/p-multigrid strategy.Our findings reveal that the proposed reinforcement learning h/p-multigrid approach significantly accelerates and improves the robustness of steady-state simulations for one-dimensional advection-diffusion and nonlinear Burgers' equations, when discretized using high-order h/p methods, on uniform and nonuniform grids. |
| format | Article |
| id | doaj-art-c987ec13c4374fc8b9ef3c258b049202 |
| institution | Kabale University |
| issn | 2590-1230 |
| language | English |
| publishDate | 2024-12-01 |
| publisher | Elsevier |
| record_format | Article |
| series | Results in Engineering |
| spelling | doaj-art-c987ec13c4374fc8b9ef3c258b0492022024-12-19T10:57:32ZengElsevierResults in Engineering2590-12302024-12-0124102949A reinforcement learning strategy to automate and accelerate h/p-multigrid solversDavid Huergo0Laura Alonso1Saumitra Joshi2Adrian Juanicotena3Gonzalo Rubio4Esteban Ferrer5ETSIAE-UPM-School of Aeronautics, Universidad Politécnica de Madrid, Plaza Cardenal Cisneros 3, E-28040 Madrid, Spain; Corresponding author.ETSIAE-UPM-School of Aeronautics, Universidad Politécnica de Madrid, Plaza Cardenal Cisneros 3, E-28040 Madrid, SpainETSIAE-UPM-School of Aeronautics, Universidad Politécnica de Madrid, Plaza Cardenal Cisneros 3, E-28040 Madrid, SpainETSIAE-UPM-School of Aeronautics, Universidad Politécnica de Madrid, Plaza Cardenal Cisneros 3, E-28040 Madrid, SpainETSIAE-UPM-School of Aeronautics, Universidad Politécnica de Madrid, Plaza Cardenal Cisneros 3, E-28040 Madrid, Spain; Center for Computational Simulation, Universidad Politécnica de Madrid, Campus de Montegancedo, Boadilla del Monte, 28660 Madrid, SpainETSIAE-UPM-School of Aeronautics, Universidad Politécnica de Madrid, Plaza Cardenal Cisneros 3, E-28040 Madrid, Spain; Center for Computational Simulation, Universidad Politécnica de Madrid, Campus de Montegancedo, Boadilla del Monte, 28660 Madrid, SpainWe explore a reinforcement learning strategy to automate and accelerate h/p-multigrid methods in high-order solvers. Multigrid methods are very efficient but require fine-tuning of numerical parameters, such as the number of smoothing sweeps per level and the correction fraction (i.e., proportion of the corrected solution that is transferred from a coarser grid to a finer grid). The objective of this paper is to use a proximal policy optimization algorithm to automatically tune the multigrid parameters and, by doing so, improve stability and efficiency of the h/p-multigrid strategy.Our findings reveal that the proposed reinforcement learning h/p-multigrid approach significantly accelerates and improves the robustness of steady-state simulations for one-dimensional advection-diffusion and nonlinear Burgers' equations, when discretized using high-order h/p methods, on uniform and nonuniform grids.http://www.sciencedirect.com/science/article/pii/S2590123024012040Reinforcement learningProximal policy optimizationPPOAdvection-diffusionBurgers' equationHigh-order flux reconstruction |
| spellingShingle | David Huergo Laura Alonso Saumitra Joshi Adrian Juanicotena Gonzalo Rubio Esteban Ferrer A reinforcement learning strategy to automate and accelerate h/p-multigrid solvers Results in Engineering Reinforcement learning Proximal policy optimization PPO Advection-diffusion Burgers' equation High-order flux reconstruction |
| title | A reinforcement learning strategy to automate and accelerate h/p-multigrid solvers |
| title_full | A reinforcement learning strategy to automate and accelerate h/p-multigrid solvers |
| title_fullStr | A reinforcement learning strategy to automate and accelerate h/p-multigrid solvers |
| title_full_unstemmed | A reinforcement learning strategy to automate and accelerate h/p-multigrid solvers |
| title_short | A reinforcement learning strategy to automate and accelerate h/p-multigrid solvers |
| title_sort | reinforcement learning strategy to automate and accelerate h p multigrid solvers |
| topic | Reinforcement learning Proximal policy optimization PPO Advection-diffusion Burgers' equation High-order flux reconstruction |
| url | http://www.sciencedirect.com/science/article/pii/S2590123024012040 |
| work_keys_str_mv | AT davidhuergo areinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT lauraalonso areinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT saumitrajoshi areinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT adrianjuanicotena areinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT gonzalorubio areinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT estebanferrer areinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT davidhuergo reinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT lauraalonso reinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT saumitrajoshi reinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT adrianjuanicotena reinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT gonzalorubio reinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers AT estebanferrer reinforcementlearningstrategytoautomateandacceleratehpmultigridsolvers |