提高遗传算法计算速度的研究

在网络规划中, 由于遗传算法的引进, 提高了网络优化的准确度和工作效率, 但是计算速度慢是遗传算法存在的一个重大缺陷, 这直接影响算法的实际应用效果。实验表明, 影响遗传算法计算速度的因素主要集中在算法搜索空间的大小、进化过程中父代染色体的选择机制、交叉运算导致的局部优化积累以及进化过程中无效染色体的处理等几个方面。相应地采用问题空间的制约机制减少搜寻空间;采用动态阈值择优选择父代染色体;采用轮盘赌法进行选择, 并加入最优保留策略, 既维持种群的多样性, 打破局部优化积累, 又保证了最优个体直接进入下一代;采用修正策略处理无效染色体, 减少了循环次数, 同时增加样本多样性避免局部收敛, 提高了网络优化的效率。

Improving Computing Speed of Genetic Algorithm

In the network planning,introducing genetic algorithm improves accuracy and efficiency of network optimization.The slow computing speed is an obvious deficiency of genetic algorithm,which influences directly application effect of the algorithm.A method to solve the problem was proposed through analyzing the reason and mechanism of influencing computing speed of genetic algorithm.Experiments prove that the method is correct and effective.According to experiment results,factors to influence computing speed of genetic algorithm mainly relates to size of search space,select mechanism of paternal chromosome in the process of evolution,partial optimization accumulation resulted from crossed computation,and handling of invalid chromosome during evolution.Accordingly,search space is decreased by adopting restrict mechanism of problem space;paternal chromosome is selected through dramatic threshold value;through adopting roulette method for selection and adding optimum reserved strategy,diversity of colony is kept to break partial optimization accumulation,the optimum individual is insured to enter the next generation directly;circular number is reduced through applying revised strategy to handle the invalid chromosome,and at the same time,diversity of samples is increased and partial convergence is avoided.Thus,the efficiency of network optimization is improved.