Massively Parallel Asynchronous Fractal ...
Document type :
Communication dans un congrès avec actes
Title :
Massively Parallel Asynchronous Fractal Optimization
Author(s) :
Firmin, Thomas [Auteur]
Université de Lille
Optimisation de grande taille et calcul large échelle [BONUS]
Talbi, El-Ghazali [Auteur]
Inria Lille - Nord Europe
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]
Université de Lille
Optimisation de grande taille et calcul large échelle [BONUS]
Université de Lille
Optimisation de grande taille et calcul large échelle [BONUS]
Talbi, El-Ghazali [Auteur]
Inria Lille - Nord Europe
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]
Université de Lille
Optimisation de grande taille et calcul large échelle [BONUS]
Conference title :
2023 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW)
City :
St. Petersburg
Country :
France
Start date of the conference :
2023-05-15
Publisher :
IEEE
English keyword(s) :
Fractal optimization
HAL domain(s) :
Informatique [cs]
English abstract : [en]
Fractal-based decomposition is a flexible framework representing a family of optimization algorithms based on a hierarchical decomposition of the search space. We have built a software called Zellij, in which we were able ...
Show more >Fractal-based decomposition is a flexible framework representing a family of optimization algorithms based on a hierarchical decomposition of the search space. We have built a software called Zellij, in which we were able to instantiate popular decomposition-based algorithms. Our goal is to tackle optimization problems characterized by computationally expensive objective functions and high dimensional search space. In this paper, we propose a generic asynchronous parallel methodology of fractal-based optimization algorithms on multi-nodes and multi-CPUs distributed environments. Experimental results show a significantly reduced computation time between the mono-threaded version and the asynchronous one. The obtained results are also analyzed according to the various search components such as tree search, exploration, and exploitation strategies.Show less >
Show more >Fractal-based decomposition is a flexible framework representing a family of optimization algorithms based on a hierarchical decomposition of the search space. We have built a software called Zellij, in which we were able to instantiate popular decomposition-based algorithms. Our goal is to tackle optimization problems characterized by computationally expensive objective functions and high dimensional search space. In this paper, we propose a generic asynchronous parallel methodology of fractal-based optimization algorithms on multi-nodes and multi-CPUs distributed environments. Experimental results show a significantly reduced computation time between the mono-threaded version and the asynchronous one. The obtained results are also analyzed according to the various search components such as tree search, exploration, and exploitation strategies.Show less >
Language :
Anglais
Peer reviewed article :
Oui
Audience :
Internationale
Popular science :
Non
ANR Project :
Collections :
Source :