Abstract: Srinvivas etc. have proposed an adaptive genetic algorithm，whose cross-probability and the probability of variation can adapt to change with the size of sufficiency.In this algorithm，cross-probability and the probability of variation of individual who have the biggest sufficiency value is zero，which makes evolution toward the possibility of partial optimal solution to increase.Therefore，an improved adaptive genetic algorithm is proposed，in which cross-probability and the probability of variation of individual who have the biggest sufficiency value is not zero.The test results indicate that this algorithm can suppress "premature"，can prevent falling into a local optimum，can enhance the rate of the population’s convergence.

Key words: genetic algorithm, adaptive genetic algorithm, premature convergence, optimal value

摘要： Srinvivas等提出一种自适应遗传算法，交叉概率与变异概率能够随着适应度大小而改变。但在这种算法中，群体中最大适应度值的个体的交叉率和变异率为零，这使得进化走向局部最优解的可能性增加。提出了一种改进的自适应遗传算法，使群体中最大适应度值的个体的交叉率和变异率不为零。实验结果表明该算法在抑制“早熟”现象，防止陷入局部最优，提高种群收敛速度方面都有明显的效果。

关键词: 遗传算法, 自适应遗传算法, 早熟, 最优值