In this paper we extend the classical Evolution Strategies by a new mechanism to adjust the step sizes. We propose an Evolution Strategy with Variance Function (ESV). The ESV can generate local and global random search procedures depending on the task. The idea of the variance function approach is presented. A performance comparison of the ESV with other published Evolutionary Algorithms for global optimization problems is made. We report about some aspects of the analysis of the used multimodal test functions which concern the complexity of the fitness landscape.
展开▼