Directional self-learning of genetic algorithm

In order to overcome the low convergence speed and prematurity of classical genetic algorithm, an improved method named directional self-learning of genetic algorithm (DSLGA) is proposed in this paper. Through the self-learning operator directional information was introduced in local search process. The search direction was guided by the false derivative of the function fitness. Using the four operators among the individuals, the best solution was updated continuously. In experiments, DSLGA was tested on 4 unconstrained benchmark problems, and the results were compared with the algorithms presented recently. It showed that DSLGA performs much better than the other algorithms both in the quality of the solutions and in the computational complexity.