Chaotic Parallel Genetic Algorithm with Variable-Scale Learning and Balancing Strategy of Ranking Individuals

Keeping balance between the diversity of population and the convergence of evolution for genetic algorithm remains a work of art. It is well known that the chaotic mapping helps to maintain good diversity for population and Baldwin effect based posterior learning promotes evolution along the right direction, thus forming chaotic parallel genetic algorithm with Baldwin learning (CPGABL). In this paper, two critical improvements are introduced into our previous works about CPGABL: first, balancing strategy of ranking individuals is adopted to guarantee first the diversity of population and then the convergence of algorithm; second, rearrangement to chaotic sequences is redesigned to maintain both good diversity and appropriate computational complexity. Performances of this enhanced CPGABL and our previous works are compared on a benchmark constrained nonlinear optimization problem.