DocumentCode :
2560473
Title :
JG2A: A Grid-enabled object-oriented framework for developing genetic algorithms
Author :
Bernal, Andrés ; Ramírez, Mauricio A. ; Castro, Harold ; Walteros, Jose L. ; Medaglia, Andrés L.
Author_Institution :
Univ. de los Andes, Bogota, Colombia
fYear :
2009
fDate :
24-24 April 2009
Firstpage :
67
Lastpage :
72
Abstract :
Java genetic algorithm (JGA) is a flexible object-oriented framework for rapid prototyping of evolutionary algorithms. Even though JGA has proven to be flexible and efficient in practice, parallelization opens new avenues to the framework. Java grid-enabled genetic algorithm (JG2A) is a new generation of JGA that exploits parallelism in genetic algorithms in two ways: first, it allows the execution in parallel of a large set of instances (instances parallelization); and second, it provides parallelization of the population evaluation (population evaluation parallelization). We illustrate instances parallelization in different parameter tuning experiments of vehicle routing and route design problems. The population evaluation parallelization is particularly useful for hard black-box optimization problems where the fitness function evaluation embeds a discrete-event or finite-element analysis simulation. JG2A can be deployed in a heterogeneous computational environment enabled by a grid based on Globus and Condor acting as the local resource manager.
Keywords :
Java; finite element analysis; genetic algorithms; grid computing; JG2A; Java genetic algorithm; evolutionary algorithms; finite-element analysis simulation; grid-enabled object-oriented framework; hard black-box optimization problems; population evaluation parallelization; vehicle routing; Computational modeling; Evolutionary computation; Finite element methods; Genetic algorithms; Java; Mesh generation; Parallel processing; Prototypes; Routing; Vehicles;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Systems and Information Engineering Design Symposium, 2009. SIEDS '09.
Conference_Location :
Charlottesville, VA
Print_ISBN :
978-1-4244-4531-8
Electronic_ISBN :
978-1-4244-4532-5
Type :
conf
DOI :
10.1109/SIEDS.2009.5166157
Filename :
5166157
Link To Document :
بازگشت