This work presents theoretical considerations and experimental studies on the working mechanisms of the differential evolution (DE) algorithm for nonlinear optimization. A constant relationship between the covariance matrix of the population at a given iteration of the algorithm, and the corresponding covariance matrix of the probability distribution for the differential mutation operator, represented by the set of all possible mutation vectors, is obtained. This relationship is then employed for analyzing the adaptive behavior of the differential evolution algorithm.
|Title of host publication||Anais do 10. Congresso Brasileiro de Inteligência Computacional|
|Number of pages||8|
|Publication status||Published - 2011|