MFGA: a GA for complex real-world optimisation problems. (1st January 2011)

• Record Type: Journal Article
• Authors: Turco, Alessandro; Kavka, Carlos
• Is Part Of: International journal of innovative computing and applications. Volume 3:Number 1(2011)
• Page Start: 31
• Page End: 41
• Publication Date: 2011-01-01
• Subjects: elitism -- genetic algorithms -- multi-objective optimisation -- SoC design; Evolutionary computation -- Periodicals; Neural networks (Computer science) -- Periodicals; Genetic programming (Computer science) -- Periodicals; Biologically-inspired computing -- Periodicals; Swarm intelligence -- Periodicals; Quantum computers -- Periodicals; 006.3
• DOI: 10.1504/IJICA.2011.037949
• Languages: English
• ISSNs: 1751-648X

Abstract

We present a multi-objective genetic algorithm called magnifying front genetic algorithm (MFGA) designed in order to treat complex real-world optimisation problems. A first source of complexity is the presence of different input variables classes (real, discrete and categorical). MFGA is able to treat appropriately each of them as well as any combination. Moreover, real-world applications often require a long time to evaluate objective values from input variables. We deal with this issue working on elitism (in order to tune properly the balance between explorative and exploitative capabilities of the algorithm) and introducing a parallel steady-state evolution scheme, which is able to use the available computing resources as much intensively as possible. We test the algorithm on two different scenarios: mathematical benchmarks and real-world applications. For the latter one we chose a problem arising in multi-processor system-on-chip (MPSoC) design, a field which is characterised by discrete and more often categorical variables.