Trade-Off Between Diversity and Convergence in Multi-objective Genetic Algorithms

Multi-objective genetic algorithms allow solving complex problems. They are often used to solve real-world problems. However, close scrutinizes of the execution of these algorithms show that they could suffer from premature convergence or diversity loss problems. This has an impact on the performance results. This paper introduces some tools for genetic algorithms to dynamically adapt their behaviors in order to avoid traps such as local optima. These tools lead to a trade-off between the exploitation and exploration steps. For this end, some quality criteria are introduced to assess solutions over generations. Thereafter, four execution modes are proposed to alternatively ensure diversity preservation and convergence. The results presented in this paper show that the use of these tools improves the overall performance of genetic algorithms.