【摘要】 分析了John Holland在其遗传算法理论中提出的反转算子在数值优化应用中的不合理性,提出了一种轮盘反转算子来克服这种不合理性,并结合新的反转算子在杂草生态学的基础上提出了一种高效的多智能体仿生算法—RMAAA算法,证明了算法的全局收敛性。14个10～10000维无约束优化问题仿真实验表明,该算法性能好于其他优秀算法,尤其对于其中8个函数,当维数为10000时该算法需要的评价次数都小于10000次,且所得解的精度较高。在求解焊接梁设计工程优化问题时,该算法能够以较少的函数评价次数得到更加精确的全局Pareto最优解,表明RMAAA算法是实际有效的。更多还原

【Abstract】 The irrationality of the inversion operator designed by John Holland was analyzed when used for numerical optimization, and a new roulette inversion operator was proposed to conquer that irrationality.Then a new artificial algorithm---RMAAA was designed integrating the roulette inversion operator.Theoretical analysis shows that RMAAA converges to the global optimum.In the first part of the experiments, 14 benchmark functions are used to test the performance of RMAAA, and the scalability of RMAAA along the problem dimension ranging 10～10000 is studied with good care.The results show that RMAAA achieves a better performance than other algorithms, and for 8 of 14 functions RMAAA only needs less than 10000 times of function evaluation to find high quality solutions even when the dimensions are increased to as high as 10000.In the following part of the experiments, RMAAA is used to solve a practical design optimization problem---minimum cost of a welded beam in both fixed search areas and dynamically expanded search area with satisfactory results.Experiments show that RMAAA is effective and practical.更多还原