A novel method to improve the performance of chaos based evolutionary algorithms

Abstract

Many methods have been proposed to improve performance of evolutionary algorithms. One of the most common methods used in these studies is the chaos theory. Chaotic systems are especially widely used in creation of the initial populations and to express optimization variables. Although, chaos theory is a good application field in the design of evolutionary algorithms, many proposals become unsatisfactory. Since using many chaotic systems do not have a uniform distribution. In this study, we analyze the weaknesses of the evolutionary algorithms based on chaotic maps. By applying statistical distribution test, we show that using chaos based pseudorandom number generators are not appropriate for pseudorandom characteristics. In order to overcome this problem, a novel method has been proposed. Statistical tests and distribution test show that proposed new generator has good pseudorandom characteristics. (C) 2015 Elsevier GmbH. All rights reserved.