Parallel Evolutionary Computation Framework for Single- and Multiobjective Optimization
Evolutionary computation is an area of computer science utilizing the mechanisms of biological evolution in computer problem solving. It is concerned with theoretical studies, design and application of stochastic optimization procedures, known as Evolutionary Algorithms (EAs). EAs have proven effective and robust in solving demanding optimization problems that are often difficult if not intractable to traditional numerical methods. They are nowadays widely applied in science, engineering, management, and other domains. However, a drawback of EAs is their computational complexity which originates from iterative population-based search of the solution space. On the other hand, processing a population of candidate solutions makes EAs amenable to parallel implementation that may result in significant calculation speedup.
This chapter presents a parallel evolutionary computation framework developed for solving numerical optimization problems with one or more objectives, and evaluates its performance on a high-dimensional optimization task from industrial practice. The chapter starts with an introduction to optimization problems. It distinguishes between single- and multiobjective optimization and reviews the concepts needed to deal with multiobjective optimization problems, such as the dominance relation and Pareto optimality. Next, EAs as a general-purpose optimization method are described, with a focus on Differential Evolution (DE) which is a particular kind of EA used in our framework. Then, parallelization of EAs is discussed in view of known parallelization types and speedup calculation. The chapter continues with an introduction to the optimization problem in industrial continuous casting, used as a test problem in this work. Afterwards, the proposed parallel evolutionary computation framework is presented. The framework is based on DE and implemented on a cluster of personal computers. It is evaluated on single- and multiobjective variants of the casting optimization problem and the results are analyzed from the perspective of the problem domain and, in particular, the achieved speedup.
Unable to display preview. Download preview PDF.
- 3.G. B. Fogel, D.W. Corne (Eds.), Evolutionary Computation in Bioinformatics, Morgan Kaufmann Publishers, Amsterdam, (2003).Google Scholar
- 6.B. Filipič, E. Laitinen, Model-based tuning of process parameters for steady-state steel casting, Informatica 29 (4) (2005) 491–496.Google Scholar
- 7.R. Buyya, High Performance Cluster Computing: Architectures and Systems, Vol. 1, Prentice Hall, Upper Saddle River, (1999).Google Scholar
- 13.T. Robič, B. Filipič, Demo: Differential evolution for multiobjective optimization, in: C. A. Coello Coello, A. Hernández Aguirre, E. Zitzler (Eds.), Conference on Evolutionary Multi-Criterion Optimization, Vol. 3410 of Lecture Notes in Computer Science, Springer, Berlin, (2005), pp. 520–533.Google Scholar
- 15.S. G. Akl, Parallel Computation: Models and Methods, Prentice Hall, Upper Saddle River, (1997).Google Scholar
- 16.E. Cantú-Paz, A survey of parallel genetic algorithms, Tech. rep., University of Illinois at Urbana-Champaign (1997).Google Scholar
- 18.M. Snir, S. Otto, S. Huss-Lederman, D. Walker, J. Dongarra, MPI – The Complete Reference, The MIT Press, Cambridge, (1996).Google Scholar
- 20.B. Filipič, T. Tušar, E. Laitinen, Preliminary numerical experiments in multiobjective optimization of a metallurgical production process, Informatica 31 (2) (2007) 233–240.Google Scholar