Restartings: A Technique to Improve Classic Genetic Algorithms Performance

Authors: Grigorios N. Beligiannis, Georgios A. Tsirogiannis, Panayotis E. Pintelas

Abstract:

In this contribution, a way to enhance the performance of the classic Genetic Algorithm is proposed. The idea of restarting a Genetic Algorithm is applied in order to obtain better knowledge of the solution space of the problem. A new operator of 'insertion' is introduced so as to exploit (utilize) the information that has already been collected before the restarting procedure. Finally, numerical experiments comparing the performance of the classic Genetic Algorithm and the Genetic Algorithm with restartings, for some well known test functions, are given.

Keywords: Genetic Algorithms, Restartings, Search space exploration, Search space exploitation.

Digital Object Identifier (DOI): doi.org/10.5281/zenodo.1078787

References:

[1] D. E. Goldberg, Genetic Algorithms in Search, Optimization and Machine Learning, Reading, Mass.: Addison-Wesley, 1989.
[2] Z. Michalewicz, Genetic Algorithms + Data Structures = Evolution Programs, 3rd ed., N. Y.: Springer-Verlag, 1996.
[3] M. Mitchell, An Introduction to Genetic Algorithms (Complex Adaptive Systems), Cambridge, Massachusetts, London, England: A Bradford Book, The MIT Press, 1998.
[4] H. A.. van der Vorst, ''Computational Methods for large Eigenvalue Problems'', in Handbook of Numerical Analysis, vol. 8, P. G. Ciarlet and J. L. Lions, Eds. Amsterdam: North-Holland (Elsevier), pp. 3-179, 2002.
[5] Y. Saad, Numerical methods for large eigenvalue problems,Manchester, UK: Manchester University Press, 1992.
[6] S. G. Petition, ''Parallel subspace method for non-Hrmitian eigen-problems on the connection machine (CM2)'', Applied Numerical Mathematics, vol.10, pp. 19-36, 1992.
[7] D. C. Sorensen, ''Implicit application of polynomial filters in a k-step Arnoldi method'', SIAM J. Matrix Anal. Applic., vol. 13(1), pp. 357-385, 1992.
[8] K. De Jong, ''An analysis of the behaviour of a class of genetic adaptive systems'', PhD thesis, University of Michigan, 1975.
[9] D. M. HimmelBlau, Applied Linear Programming, McGraw-Hill, 1972.
[10] GAlib - A C++ Library of Genetic Algorithm Components, Matthew Wall, Massachusetts Institute of Technology (MIT). Available: http://lancet.mit.edu/ga/