An Improvement to Genetic Algorithms for Multimodal Optimization in Noisy Environments: Re-evaluation of All Individuals per Generation

Optimization in noisy environments is regard as a favorite application domains of genetic algorithms. Different methods for reducing the influence of noise are presented and discussed. A new fitness evaluation method is proposed that reevaluates all survival individuals each generation. Compared with re-sampling and population sizing, the new evaluation approach shows higher probability of searching to the global extremum area and precision of convergence. These results demonstrate that the proposed method is effective for reducing noise effects.