An Adaptive Differential Evolution Algorithm Based on Fuzzy Modeling

The appropriate parameter setting can substantially determine the performance of differential evolution (DE), so parameter design is a very crucial and challenging task in DE. In response to the realistic demands, a novel adjust strategy for adaptive parameter is developed for DE in this paper. By way of the strategy of fuzzy modeling, the phases of optimization are designed as follows, i.e., exploration, exploitation and convergence. The adaptive adjust of F and CR, the control parameters, is determined by the phases of optimization. Meanwhile, an auxiliary movement technique is designed for the convergence population. This technique will help the best individual to avoid the risk of falling into the potential local optima. The proposed algorithm, namely FMDE/rand/1, has been assessed under eight unimodal and multimodal benchmark functions. Results from experiments illustrate that the proposed FMDE/rand/1 is a promising optimization algorithm which will greatly enhance the performance on effectiveness and dynamic.