基于混合优化算法的遗传算法参数设定研究

有限计算量条件下遗传算法的理论收敛条件难以完全满足,参数选择的恰当与否直接影响到算法性能的发挥。针对这一情况,在分析现有参数设定方法的基础上,将遗传算法参数设定问题描述为随机优化问题,并提出一种解决该问题的新的混合优化算法,即基于序优化的巢分区算法。该算法将序优化思想融入巢分区算法的局部搜索过程,大大提高了局部搜索效率,而巢分区的算法框架则保证了算法的全局收敛性。以典型旅行商问题为算例的仿真结果验证了该方法的高效性与可靠性。

Study on parameter setting for genetic algorithm based on hybrid optimization method

Rigorous theoretical convergent conditions of the Genetic Algorithm(GA) can not be met with limited computing budget constraints,and thereby the fitness of selected parameters influences the performance of algorithms greatly and directly.After analyzing existing literature on this problem,the parameter setting of GA described as a stochastic optimization problem is introduced and a new hybrid optimization algorithm named Ordinal Optimization based Nested Partitions algorithm is proposed.The hybrid algorithm incorporates Ordinal Optimization into the local search process of the Nested Partitions(NP) method,which improves the efficiency of NP greatly.And the NP method can guarantee the global convergence of new algorithm.Simulation results for several typical traveling salesman problems indicate the validity and reliability of the proposed method.