A Crossover Optimization Method based on Correlation Coefficient in Differential Evolution

Optimiz ation is to carry out characteristic distribution according to the dependencies in complex variables, mainly in the form of thin ellipses. To achieve the best effect in the dependency relationship, it is necessary to slide the long axis of the ellipse and change the variables at the same time. In addition, under the same distribution state, the optimal solution will also appear when the set of search points is far away in the variable separation problem. In differential evolution, it is difficult to cross a specific genetic gene at the same time because it determines whether their variables cross with the same probability. In this paper, to test this shape, a method of the searching point correlation coefficient is proposed The correlation matrix is obtained according to the distribution of search points, and the genes with a strong correlation coefficient are grouped into groups, which are crossed or not crossed at the same time. This method is introduced into the representative method of differential evolution JADE to optimize the basic problems. The experimental results of this method are verified by comparing the performance.