基于位变异防止遗传算法过早收敛的算法

An Genetic Algorithm Based on Bit Mutation to Avoid Premature Convergence

针对简单遗传算法具有过早收敛的缺点,本文提出了一种基于位变异的防止遗传算法过早收敛的算法。该算法通过种群熵来判断过早收敛的发生。当发生过早收敛时,在单调系数的指导下进行有针对性的位变异,从局部最优解的范围内摆脱出来,算法重新具有进化能力。用数值优化中的三个典型的测试函数对该算法进行测试,仿真实验结果显示该算法可有效地提高算法的全局搜索能力。

In this paper an improved genetic algorithm is presented to solve the problem of genetic algorithm premature convergence, which can forecast the occurrence of premature convergence by using colony entropy. The algorithm can break out of the trap of the partial best value by means of the mutation directed by the monotonic coefficient of population. When the premature phenomena occurs, the algorithm restores the ability of evolution. The algorithm is tested by three typical test functions of numerical optimization, and the experimental results show that the improved algorithm has greater probability of convergence.