蚁群协同模式搜索算法及其收敛性分析

Ant colony pattern search algorithms and their convergence

提出了一种解决无约束连续空间优化问题的蚁群协同模式搜索算法.该算法通过目标函数值启发式信息素引导群体进行区域搜索,而每个个体的模式搜索为算法提供进一步的局部搜索,其搜索结果以信息素融合的方式进行信息共享,为下一次的区域搜索提供依据.通过随机模式搜索算法理论得出了算法的收敛性定理.详细的测试结果体现算法的涌现智能特征,与其他算法的比较结果说明了算法的有效性及群体协同的优势.

A class of ant colony pattern search algorithms （ACPSAs） are designed for the optimization of multimodal functions in continuous space. ACPSAs guide the individuals to perform region searches by objective function heuristic pheromone. Further local searches are handled by pattern searches of individuals, then the search results are shared with pheromone fusion, providing the basis for the region searches in the next iteration. The probabilistic convergence theories of ACPSAs are also given by stochastic pattern search algorithm theory. APCSAs present interesting emergent properties as shown by some analytical test functions. Finally, the comparison results with typical stochastic optimization algorithms show the effectiveness of the algorithms and the advantage in swarm cooperation.