A novel evolutionary algorithm with fast convergence

This paper proposes a novel evolutionary algorithm of which convergence speed is improved from the evolutionary programming without decreasing the diversity. The proposed algorithm has two mutation operators according to the evolution conditions, respectively. One is a direction operator according to the value of cost function. The other is a Gaussian perturbation whose mean is zero. Additionally, a variable “age” is used to enhance the diversity of the search and prevents individuals from remaining in the local minima. An offspring is selected if it wins in competition with its parent. Comparison between the proposed algorithm and evolutionary programming is carried out for the eight test functions to show the effectiveness of the proposed one