基于殖民地位置信息的自适应帝国主义竞争算法

Adaptive imperialist competitive algorithm based on colony location information

在元启发式算法中,问题优化的质量取决于算法参数的精准设定,而参数的设定通常没有一个确定的标准,重复实验又耗时耗力。偏离角度是帝国主义竞争算法(ICA)中的一个重要参数,盲目选取可能会导致算法陷入局部最优解,出现早熟收敛现象。为了克服该缺陷,提出一种基于殖民地位置信息概率密度函数的自适应帝国主义竞争算法。在帝国主义者同化殖民地过程中,根据殖民地密度函数动态地调整其向帝国主义者运动的偏离角度,提高殖民地脱离局部最优解探寻全局最优解的能力。另一方面,帝国主义者通过及时掠取殖民地位置中的有用信息降低自己的成本函数,延伸了原算法中殖民地与帝国主义者位置互换的步骤,加快了算法的收敛速度。在一系列的基准函数测试中,所提算法在收敛速度和优化性能上均优于原ICA和另外几种经典的遗传算法。

In the meta-heuristic algorithm,the quality of problem optimization depends on the setting precision of algorithm parameters,but there is no definitive standard for parameter setting,and repeated trials consume time and strength. Deviation angle is an important parameter in imperialist competitive algorithm(ICA),and blind selection tends to make the algorithm trapped into local optimal solution,resulting in early convergence. To overcome this defect,an adaptive imperialist competitive algorithm based on the probability density function of colony location information is proposed. In the colony assimilation process of imperialists,the deviation angle that the colony moves to the imperialist is dynamically adjusted according to the density function of the colony so as to improve the colony′s capability of separating from the local optimal solution and seeking for the global optimal solution. The imperialists reduce their cost functions by timely capturing useful colony location information,which extends the procedures of location interchange between colonies and imperialists in the original algorithm,and speeds up the convergence of the algorithm. A series of benchmark function tests were carried out. The results show that the proposed algorithm has advantage in convergence speed and optimization performance in comparison to the original ICA and several other classical genetic algorithms.