An Efficient Improved Differential Evolution Algorithm

Differential evolution (DE) algorithm is a promising global optimization approach, but its control parameters are sensitive to some difficult problems, and they must be adjusted artificially for different problems some times, which is really a time consuming work. In this paper, we present a new version of DE with self-adaptive control parameters. We call the new version efficient improved differential evolution (EIDE). The EIDE modifies scale factor by using a uniform distribution, and modifies crossover rate by using a linear increasing strategy. Both strategies can avoid guessing the appropriate values for scale factor and crossover rate, and save the regulating time of the two parameters. Based on two groups of experiments, the EIDE has shown better convergence and stability than the other three DE algorithms in most cases.