Abstract
This paper proposes a novel adaptation scheme for Differential Evolution (DE) frameworks. The proposed algorithm, namely Estimation Distribution Differential Evolution (EDDE), is based on a DE structure and employs randomized scale factor ad crossover rate values. These values are sampled from truncated Gaussian probability distribution functions. These probability functions adaptively vary during the optimization process. At the beginning of the optimization the truncated Gaussian functions are characterized by a large standard deviation values and thus are similar to uniform distributions. During the later stages of the evolution, the probability functions progressively adapt to the most promising values attempting to detect the optimal working conditions of the algorithm. The performance offered by the proposed algorithm has been compared with those given by three modern DE based algorithms which represent the state-of-the-art in DE. Numerical results show that the proposed EDDE, despite its simplicity, is competitive with the other algorithms and in many cases displays a very good performance in terms of both final solution detected and convergence speed.