Intelligent control of dynamic population size for evolutionary algorithms

We are presenting an innovative method for improving the performance of single objective (SO) evolutionary algorithms (EAs). It consists of an intelligent method based in human expertise to establish a fuzzy inference system with the purpose of making more efficient the exploration and exploitation of the landscape by increasing or diminishing the amount of individuals through generations. It is very common that human experts disagree in a major or a minor part of the knowledge hence information can be somewhat contradictory and doubtful. The proposed method uses a novel concept called Mediative Fuzzy Logic (MFL) for handling doubtful and contradictory information from experts with the aim of calculating the appropriated amount of individual to create and/or to eliminate. We are presenting results that outperform significantly in precision and in time the Genetic Algorithm of the Matlab's Toolbox.