Chaotic genetic algorithms

Abstract

The convergence properties of Genetic Algorithms are closely connected to the random sequence applied on the genetic operators during a run. As chaos and random sequence share the property of long-term unpredictable irregular behaviour and broadband spectrum, we replaced the random sequence generators by chaos generators, without affecting the original operator definitions in genetic algorithms. Three chaos models were used, the Logistic map, the Tent map and the Chua’s circuit. Both single objective functions and multi objective functions are tested in our experiments. The results show that best performance of convergence is obtained by using chaos generators.