Phases-based dynamic genetic strategies for genetic algorithms

This paper focuses on the study of dynamic genetic strategies for adaptive parameters control in genetic algorithms (GA). The parameter space of GA is defined, and different adaptation approaches are compared. The currently adopted adaptive strategies make use of the crossover and mutation probabilities at individual level or component level, which, in the total process of evolution operations, can not fully exploiting the genetic information. Considering the dynamic population searching in the evolution process, a three phases-based adaptive genetic strategy is formulated, so that GA can be equipped simultaneously with the capabilities in both exploration and exploitation. It is then applied to the optimization of test functions, and the results reveal its efficiency and effectiveness.