Accelerating convergence towards the optimal pareto front

Evolutionary algorithms have been very popular optimization methods for a wide variety of applications. However, in spite of their advantages, their computational cost is still a prohibitive factor in certain real-world applications involving expensive (computationally speaking) fitness function evaluations. In this paper, we depart from the observation that nature's survival of the fittest is not about exact measures of fitness; rather it is about rankings among competing peers. Thus, by exploiting this natural tolerance for imprecision, we propose here a new, fuzzy granules-based approach for reducing the number of necessary function calls involving time consuming real-world problems. Our proposed approach is compared with respect to the standard NSGA-II, using the Set Coverage, Hypervolume and Generational Distance performance measures. Our results indicate that our proposed approach is a very promising alternative for dealing with multi-objective optimization problems involving expensive fitness function evaluations.