Asynchronous Master-Slave Parallelization of Differential Evolution for Multi-Objective Optimization

被引:46
作者
Depolli, Matjaz [1 ]
Trobec, Roman [1 ]
Filipic, Bogdan [2 ]
机构
[1] Jozef Stefan Inst, Dept Commun Syst, SL-1000 Ljubljana, Slovenia
[2] Jozef Stefan Inst, Dept Intelligent Syst, SL-1000 Ljubljana, Slovenia
基金
欧盟地平线“2020”;
关键词
Multi-objective optimization; evolutionary algorithms; differential evolution; parallelization; distributed computing; speedup; selection lag; GENETIC ALGORITHMS; MODEL;
D O I
10.1162/EVCO_a_00076
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In this paper, we present AMS-DEMO, an asynchronous master-slave implementation of DEMO, an evolutionary algorithm for multi-objective optimization. AMS-DEMO was designed for solving time-intensive problems efficiently on both homogeneous and heterogeneous parallel computer architectures. The algorithm is used as a test case for the asynchronous master-slave parallelization of multi-objective optimization that has not yet been thoroughly investigated. Selection lag is identified as the key property of the parallelization method, which explains how its behavior depends on the type of computer architecture and the number of processors. It is arrived at analytically and from the empirical results. AMS-DEMO is tested on a benchmark problem and a time-intensive industrial optimization problem, on homogeneous and heterogeneous parallel setups, providing performance results for the algorithm and an insight into the parallelization method. A comparison is also performed between AMS-DEMO and generational master-slave DEMO to demonstrate how the asynchronous parallelization method enhances the algorithm and what benefits it brings compared to the synchronous method.
引用
收藏
页码:261 / 291
页数:31
相关论文
共 47 条
[1]  
Akl S.G., 1997, Parallel Computation: Models and Methods
[2]   Parallel evolutionary algorithms can achieve super-linear performance [J].
Alba, E .
INFORMATION PROCESSING LETTERS, 2002, 82 (01) :7-13
[3]   Improving flexibility and efficiency by adding parallelism to genetic algorithms [J].
Alba, E ;
Troya, JM .
STATISTICS AND COMPUTING, 2002, 12 (02) :91-114
[4]  
Auger A, 2009, FOGA'09: PROCEEDINGS OF THE 10TH ACM SIGRVO CONFERENCE ON FOUNDATIONS OF GENETIC ALGORITHMS, P87
[5]   Grid'5000:: A large scale and highly reconfigurable experimental grid testbed [J].
Bolze, Raphael ;
Cappello, Franck ;
Caron, Eddy ;
Dayde, Michel ;
Desprez, Frederic ;
Jeannot, Emmanuel ;
Jegou, Yvon ;
Lanteri, Stephane ;
Leduc, Julien ;
Melab, Noredine ;
Mornet, Guillaume ;
Namyst, Raymond ;
Primet, Pascale ;
Quetier, Benjamin ;
Richard, Olivier ;
Talbi, El-Ghazali ;
Touche, Irea .
INTERNATIONAL JOURNAL OF HIGH PERFORMANCE COMPUTING APPLICATIONS, 2006, 20 (04) :481-494
[6]  
Cantu-Paz E, 1997, TECHNICAL REPORT
[7]  
Deb K, 2004, ADV INFO KNOW PROC, P105
[8]   A fast and elitist multiobjective genetic algorithm: NSGA-II [J].
Deb, K ;
Pratap, A ;
Agarwal, S ;
Meyarivan, T .
IEEE TRANSACTIONS ON EVOLUTIONARY COMPUTATION, 2002, 6 (02) :182-197
[9]   Parallel evolutionary optimization of multibody systems with application to railway dynamics [J].
Eberhard, P ;
Dignath, F ;
Kübler, L .
MULTIBODY SYSTEM DYNAMICS, 2003, 9 (02) :143-164
[10]  
Efron B, 1982, SIAM, DOI DOI 10.1093/ICESJMS/FSQ170