Aalto University Schools of Technology - electronic academic dissertations - http://otalib.aalto.fi/fi/kokoelmat_tiedonhaku/e-julkaisut/vaitoskirjat/
Aalto

Methods for Improving Reliability of Evolutionary Computation Algorithms and Accelerating Problem Solving

Jarno Martikainen

Dissertation for the degree of Doctor of Science in Technology to be presented with due permission of the Department of Electrical and Communications Engineering for public examination and debate in Auditorium S3 at Helsinki University of Technology (Espoo, Finland) on the 8th of December, 2006, at 12 noon.

Overview in PDF format (ISBN 951-22-8524-X)   [4315 KB]
Dissertation is also available in print (ISBN 951-22-8523-1)

Abstract

This dissertation deals with improving the reliability of evolutionary computation algorithms and accelerating problem-solving in optimization problems. Evolutionary algorithms have proven their value in difficult optimization problems that are not usually solvable in decent time using conventional optimization methods. However, evolutionary computation methods still suffer from problems related especially to premature convergence and the lengthy run times of the algorithms. In addition, the field of evolutionary computation does not commonly use the widely accepted practices for the comprehensive statistical comparison of two different evolutionary algorithms.

This dissertation aims at improving the process of using evolutionary computation in complex optimization problems from three perspectives. First, new algorithms are proposed for demanding optimization tasks. These algorithms rely on two perspectives, using a new multipopulation approach to enable appropriate conditions for candidate solutions to evolve and fusing evolutionary algorithms with other soft computing technologies, such as fuzzy logic, in a new way. Second, this dissertation discusses a method for reducing the computational time taken to evaluate a computationally demanding objective function value using neural network-based approximations. Third, a statistical method for comparing the results produced by two different evolutionary algorithms is illustrated. This method, relying on bootstrap resampling-based multiple hypothesis testing, is known outside the field of evolutionary computation, but has not been used within the evolutionary computing community. This dissertation illustrates the use of the statistical scheme and studies the parameters affecting the interpretation of its results.

The improvements to evolutionary algorithms this dissertation proposes have been proven to be beneficial by extensive testing. The proposed algorithms and the means to reduce the time required by the objective function evaluation have shown an increase in performance when compared to the reference algorithms. This dissertation also aims at awakening discussion related to the proper use of statistics in the field of evolutionary computation.

This thesis consists of an overview and of the following 8 publications:

  1. J. Martikainen and S. J. Ovaska, Designing multiplicative general parameter filters using adaptive genetic algorithms, in Proceedings of the Genetic and Evolutionary Computation Conference, Seattle, WA, 2004, pp. 1162-1167. © 2004 Springer Science+Business Media. By permission.
  2. J. Martikainen and S. J. Ovaska, Designing multiplicative general parameter filters using multipopulation genetic algorithm, in Proceedings of the 6th Nordic Signal Processing Symposium, Espoo, Finland, 2004, pp. 25-28. © 2004 IEEE. By permission.
  3. J. Martikainen and S. J. Ovaska, Fitness function approximation by neural networks in the optimization of MGP-FIR filters, in Proceedings of the 2006 IEEE Mountain Workshop on Adaptive and Learning Systems, Logan, UT, 2006, pp. 231-236. © 2006 IEEE. By permission.
  4. J. Martikainen and S. J. Ovaska, Hierarchical two-population genetic algorithm, International Journal of Computational Intelligence Research, vol. 2, no. 4, 2006, in press. © 2006 Research India Publications. By permission.
  5. J. Martikainen and S. J. Ovaska, Optimizing dynamical fuzzy systems using aging evolution strategies, in Proceedings of the 9th IASTED International Conference on Artificial Intelligence and Soft Computing, Benidorm, Spain, 2005, pp. 5-10. © 2005 ACTA Press. By permission.
  6. J. Martikainen and S. J. Ovaska, Using fuzzy evolutionary programming to solve traveling salesman problems, in Proceedings of the 9th IASTED International Conference on Artificial Intelligence and Soft Computing, Benidorm, Spain, 2005, pp. 49-54. © 2005 ACTA Press. By permission.
  7. D. Shilane, J. Martikainen, S. Dudoit, and S. J. Ovaska, A general framework for statistical performance comparison of evolutionary computation algorithms, in Proceedings of the IASTED International Conference on Artificial Intelligence and Applications, Innsbruck, Austria, 2006, pp. 7-12. © 2006 ACTA Press. By permission.
  8. J. Martikainen and S. J. Ovaska, Comparison of a fuzzy EP algorithm and an AIS in dynamic optimization tasks, in Proceedings of the 2006 IEEE Mountain Workshop on Adaptive and Learning Systems, Logan, UT, 2006, pp. 7-12. © 2006 IEEE. By permission.

Errata of publication 2

Keywords: evolutionary computation, hybrid algorithm, optimization, statistical comparison

This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.

© 2006 Helsinki University of Technology

Last update 2011-05-26