并行蚁群算法中的自适应交流策略(英文)

Adaptive Exchanging Strategies in Parallel Ant Colony Algorithm

提出了并行蚁群算法中处理机间信息交流的两种策略,使得各处理机能够自适应地选择其他处理机以进行信息交换和相应信息素的全局更新.还提出了一种确定处理机之间进行信息交流的时间的策略,可以根据解的分布情况自适应地确定信息交流的时间,以取得全局收敛速度和解的多样性之间的平衡.在算法每一次信息交换后,采用自适应的更新策略,根据信息素的均匀度进行信息素的更新,从而避免了早熟和局部收敛.在MPP处理机曙光2000上对TSP问题的实验结果,表明了基于该自适应信息交换策略的并行蚁群算法比其他算法具有更好的收敛性、更高的加速比和效率.

Two strategies for information exchange between processors in parallel ant colony algorithm are presented. Theses strategies can make each processor choose other processors to communicate and to update the pheromone adaptively. A strategy is also presented to adjust the time interval of information exchange adaptively according to the distribution of the solutions so as to keep balance between the convergence speed and the diversity of the solutions. The adaptive parallel ant colony algorithm （APACA） based on these strategies adaptively updates the pheromone according to the equilibrium of the pheromone distribution in each information exchange so as to avoid the precocity and local convergence. These strategies are applied to the traveling salesman problem on the massive parallel processors （MPP） Dawn 2000. Experimental results show that the algorithm has higher convergence speed, speedup and efficiency than other parallel ant algorithms.