 |
    |
Actuator Noise in Recombinant Evolution Strategies on General Quadratic Fitness Models*
Hans-Georg Beyer
Department of Computer Science XI, University of Dortmund, D-44221 Dortmund, Germanyhans-georg.beyer@cs.uni-dortmund.de
Abstract. This paper addresses the influence of actuator noise on the steady state behavior of multirecombinant evolution strategies (ES) on general quadratic fitness functions. Actuator noise degrades the ES’s ability to locate the global optimizer. After a certain transient time the ES approaches a steady state behavior characterized by an expected fitness deviation from the global optimum. This expected value is calculated and the predictions are compared with ES runs on quadratic test functions.
*This work was supported by the Deutsche Forschungsgemeinschaft (DFG) as part of the Collaborative Research Center (SFB) 531.
LNCS 3102, p. 654 ff.
lncs@springer.de
© Springer-Verlag Berlin Heidelberg 2004