Analysis the Impact of Genetic Operators in Evolution Strategies

Genetic operators play an important role in Evolution Strategies (ES). There are two important issues in the evolution process of the genetic search: exploration and exploitation. We analyze the impact of the genetic operators in ES. The Classical Evolution Strategies (CES) relies on Gaussian mutation, whereas Fast Evolution Strategies (FES) selects Cauchy distribution as the primary mutation operator. The basic genetic operators of ES, as well as their performances on a number of benchmark problems are analyzed. In this paper, we propose an Improve FES (IFES) in which the basic idea is to mix different search biases of Cauchy and Gaussian mutations. IFES generates two offspring from each parent, one is then chosen as the offspring. Extensive empirical studies have been carried out to evaluate the performances of IFES, FES and CES. From the experimental results on four widely used test functions, we can show that IFES outperforms both FES and CES. Finally we conclude that it is important to strike a balance between exploration and exploitation.