Title :
The utility of scale factor adaptation in differential evolution
Author_Institution :
Dept. of Inf. Sci., Univ. of Otago, Dunedin, New Zealand
Abstract :
Differential evolution (DE) is a simple, yet often effective, approach to real-parameter optimisation. Despite this simplicity, successfully applying DE to a specific problem typically requires careful calibration of several parameters. To simplify the application of DE to problems, researchers have turned to self-adaptive methods, which attempt to learn the `ideal´ DE configuration during the evolutionary process. This paper investigates one particular type of self-adaptive DE, self-adaptive neighbourhood search differential evolution (SaNSDE), with a particular focus on the adaptation of the scale factor generation operators. The results presented in this paper suggest that SaNSDE´s scale factor generation method may be replaced with a simpler, non-adaptive approach without degrading search performance.
Keywords :
evolutionary computation; real parameter optimisation; scale factor adaptation; scale factor generation operators; self adaptive neighbourhood search differential evolution; Gaussian distribution; Generators; Histograms; Optimization; Programming; Radio access networks; Search methods;
Conference_Titel :
Evolutionary Computation (CEC), 2010 IEEE Congress on
Conference_Location :
Barcelona
Print_ISBN :
978-1-4244-6909-3
DOI :
10.1109/CEC.2010.5586480