Simulation research based on a self-adaptive genetic algorithm

Crossover probability P_c and mutation probability P_m are important parameters of genetic algorithm. Self-adaptive genetic algorithm can reach good balance between convergence speed and global optimum by adjusting P_c and P_m adaptively according to the fitness values difference among individuals. But it is not suitable to the early period of the evolutionary process. The improved self-adaptive GA proposed by this paper can avoid this drawback. And this paper trains a neural network by using the three algorithms respectively. Simulation results show that the improved self-adaptive genetic algorithm is optimal.