There is ALife beyond convergence: using a dual sharing to adapt in time dependent optimization

Using a time dependent optimization problem as a benchmark we propose to improve the genetic algorithm's adaptiveness by an appropriate diversity management dynamics. The so-called Dual Sharing GA is an extension of a former algorithm based on the use of a genotype to phenotype mapping. We experimentally evaluate the benefits over similar approaches to diversity management. We also analyze more precisely the dynamics exhibited at both phenotypic and genotypic levels in comparison with our previous approach.