Advertisement

Are evolutionary algorithms improved by large mutations?

  • C. Kappler
Modifications and Extensions of Evolutionary Algorithms Genetic Operators and Problem Representation
Part of the Lecture Notes in Computer Science book series (LNCS, volume 1141)

Abstract

When optimizing with evolutionary algorithms in a continuous search space, mutations usually are distributed according to a Gaussian. A Gaussian distribution decays exponentially, i. e. very large mutations are highly unlikely. This bears the risk of the optimization getting caught in local extrema. A more slowly decaying distribution, e. g. a Cauchy distribution, may circumvent this problem. A Cauchy distribution allows for rare large mutations. In this paper the performance of Gaussian and Cauchy distributed mutations, in particular their robustness and rate of progress, are compared analytically and numerically in a number of examples. It turns out that, in one dimension, an algorithm working with Cauchy distributed mutations is both more robust and faster. This result cannot easily be generalized to higher dimensions, where the additional problem of finding the right direction for leaving a saddle point appears. The analysis of a simple two dimensional problem does not yet allow to draw final conclusions concerning which kind of mutations, if any, is preferable in higher dimensions.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

References

  1. 1.
    I. Rechenberg, “Evolutionsstrategie, Optimierung technischer Systeme nach Prinzipien der biologischen Evolution”, Frommann-Holzboog, Stuttgart (1973)Google Scholar
  2. 2.
    I. Rechenberg, “Evolutionsstrategie '94”, Vol. 1 of Werkstatt Bionik und Evolution-stechnik, Frommann-Holzboog, Stuttgart (1994)Google Scholar
  3. 3.
    H.-P. Schwefel, “Evolution and optimum seeking”, John Wiley & Sons Inc. (1995)Google Scholar
  4. 4.
    J.-P. Bouchard and A. Georges, “Anomalous diffusion in disordered media: statistical mechanics, models and physical applications”, Phys. Rep. 195, 127 (1990)CrossRefMathSciNetGoogle Scholar
  5. 5.
    H. Szu and R. Hartley, “Fast simulated annealing”, Phys. Lett. A 122, 157 (1987)CrossRefGoogle Scholar
  6. 6.
    L. Ingber “Very fast simulated re-annealing”, J. Math. Comp. Modelling 12, 967 (1989)CrossRefGoogle Scholar
  7. 7.
    An evolutionary algorithm without cross over is characterized by (Μ 355-01 λ), where Μ is the number of parents, and λ the number of children in each generation. In a (Μ, λ) strategy, the Μ parents of the next generation are the Μ best of the λ children. In a (Μ+λ) strategy, the parents of the next generation are the Μ best of both parents and children of the last generation. See also Refs. [1, 2, 3].Google Scholar
  8. 8.
    C. W. Gardiner “Handbook of stochastic methods”, 2nd ed., Springer (1994)Google Scholar
  9. 9.
    M. Conrad, W. Ebeling “M. V. Volkenstein, evolutionary thinking and the structure of fitness landscapes”, BioSystems 27, 125 (1992)CrossRefPubMedGoogle Scholar
  10. 10.
    Z. Michalewicz “Genetic Algorithms + Data Structures=Evolution Programs”, Springer (1992).Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 1996

Authors and Affiliations

  • C. Kappler
    • 1
  1. 1.Institut für theoretische PhysikHumboldt Universität zu BerlinBerlinGermany

Personalised recommendations