Representation and self-adaption in genetic algorithms
Robert Hinterding · Victoria University Research Repository (Victoria University) · 1996
Representation and reproduction operators are important issues in Genetic Algorithms(GAs).When optimising numerical functions some researchers advocate using floating point representation instead of bit-string representation .Floating point representation is also used in Evolutionary Strategies(ESs) and Evolutionary Programming(EP) .We show that it is not the representation that is responsible for the improved performance, but the mutation operator.By using Gaussian mutation with bit-string representation we improve the performance of GAs.We then introduce self-adaption which raises the performance to that of ESs.We also show that keeping bit-string representation can have advantages to the efficiency of GAs .