Ignoble Trails - Where Crossover Is Provably Harmful

  • J. Neal Richter
  • Alden Wright
  • John Paxton
Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 5199)

Abstract

Beginning with the early days of the genetic algorithm and the schema theorem it has often been argued that the crossover operator is the more important genetic operator. The early Royal Road functions were put forth as an example where crossover would excel, yet mutation based EAs were subsequently shown to experimentally outperform GAs with crossover on these functions. Recently several new Royal Roads have been introduced and proved to require expected polynomial time for GAs with crossover, while needing exponential time to optimize for mutation-only EAs. This paper does the converse, showing proofs that GAs with crossover require exponential optimization time on new Ignoble Trail functions while mutation based EAs optimize them efficiently.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    Mitchell, M., Forrest, S., Holland, J.H.: The royal road for genetic algorithms: Fitness landscapes and GA performance. In: Varela, F.J., Bourgine, P. (eds.) Proceedings of the First European Conference on Artificial Life. Toward a Practice of Autonomous Systems, pp. 243–254. MIT Press, Cambridge (1992)Google Scholar
  2. 2.
    Forrest, S., Mitchell, M.: Relative building-block fitness and the building-block hypothesis. In: Whitley, L.D. (ed.) Foundations of genetic algorithms 2, pp. 109–126. Morgan Kaufmann, San Mateo (1993)Google Scholar
  3. 3.
    Mitchell, M., Holland, J.H., Forrest, S.: When will a genetic algorithm outperform hill climbing. In: Cowan, J.D., Tesauro, G., Alspector, J. (eds.) NIPS, pp. 51–58. Morgan Kaufmann, San Francisco (1993)Google Scholar
  4. 4.
    Jansen, T., Wegener, I.: Real royal road functions–where crossover provably is essential. Discrete Applied Mathematics 149(1-3), 111–125 (2005)CrossRefMATHMathSciNetGoogle Scholar
  5. 5.
    Poli, R., Wright, A.H., McPhee, N., Langdon, W.: Emergent behaviour, population-based search and low-pass filtering. In: Proceedings of the Congress on Evolutionary Computation (CEC) 2006, pp. 88–95. IEEE Press, Los Alamitos (2006)Google Scholar
  6. 6.
    Deb, K., Goldberg, D.E.: Analyzing deception in trap functions. In: Whitley, L.D. (ed.) FOGA, pp. 93–108. Morgan Kaufmann, San Francisco (1992)Google Scholar
  7. 7.
    Watson, R.A.: Analysis of recombinative algorithms on a non-separable building-block problem. In: Foundations of Genetic Algorithms 6, pp. 69–89. Morgan Kaufmann, San Francisco (2001)CrossRefGoogle Scholar
  8. 8.
    Watson, R.A., Pollack, J.B.: Hierarchically consistent test problems for genetic algorithms. In: Angeline, P.J., Michalewicz, Z., Schoenauer, M., Yao, X., Zalzala, A. (eds.) Proceedings of the Congress on Evolutionary Computation, Mayflower Hotel, Washington D.C., USA, vol. 2, pp. 6–9. IEEE Press, Los Alamitos (1999)Google Scholar
  9. 9.
    Dietzfelbinger, M., Naudts, B., Hoyweghen, C.V., Wegener, I.: The analysis of a recombinative hill-climber on h-iff. IEEE Trans. Evolutionary Computation 7(5), 417–423 (2003)CrossRefGoogle Scholar
  10. 10.
    Jansen, T., Wegener, I.: The analysis of evolutionary algorithms - a proof that crossover really can help. Algorithmica 34(1), 47–66 (2002)CrossRefMATHMathSciNetGoogle Scholar
  11. 11.
    Storch, T., Wegener, I.: Real royal road functions for constant population size. Theor. Comput. Sci. 320(1), 123–134 (2004)CrossRefMATHMathSciNetGoogle Scholar
  12. 12.
    Fischer, S., Wegener, I.: The one-dimensional ising model: Mutation versus recombination. Theor. Comput. Sci. 344(2-3), 208–225 (2005)CrossRefMATHMathSciNetGoogle Scholar
  13. 13.
    Sudholt, D.: Crossover is provably essential for the ising model on trees. In: GECCO 2005: Proceedings of the 2005 conference on Genetic and evolutionary computation, pp. 1161–1167. ACM, New York (2005)Google Scholar
  14. 14.
    Doerr, B., Happ, E., Klein, C.: Crossover is provably useful in evolutionary computation. In: GECCO 2008: Proceedings of the 2008 conference on Genetic and evolutionary computation (to appear, 2008)Google Scholar
  15. 15.
    Davis, L.: Handbook of Genetic Algorithms. Van Nostrand Reinhold, New York (1991)Google Scholar
  16. 16.
    Rudolph, G.: Convergence Properties of Evolutionary Algorithms. Dr. Kovac, Hamburg (1997)Google Scholar
  17. 17.
    Motwani, R., Raghavan, P.: Randomized algorithms. Cambridge University Press, New York (1995)CrossRefMATHGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2008

Authors and Affiliations

  • J. Neal Richter
    • 1
  • Alden Wright
    • 2
  • John Paxton
    • 1
  1. 1.Montana State UniversityUSA
  2. 2.University of MontanaUSA

Personalised recommendations