A Hybrid Method of Evolutionary Algorithms and Gradient Search

This paper proposes a hybrid evolutionary algorithm. It is based on a normal evolutionary algorithm and modified with gradient search technique. The gradient individual, which is the best individual of population, propagates to the next generation’s one using gradient information. Other individuals except the best individual are distributed symmetrically near the best individual for gradient calculation. The central difference method is used for the gradient calculation and BFGS algorithm for Hessian. The gradient estimation accuracy and the overall performance of the proposed method are tested with numerical examples