分支定界算法的分布并行化研究.pdfVIP

第 26卷第 3期 计算机应用 Vol. 26 No. 3 2006年 3月 Computer pplications Mar. 2006 : 1001- 9081( 2006) 03- 0723- 04 李一明,李 毅, 周明天 (电子科技大学计算机科学与工程学院, 四川成都 610054) ( icyore@ sohu. com) :介绍了一种专用于计算分 定界算法的机群计算平台, 其中所使用的分布并行策略减少 了分 定界算法计算时间复杂度, 减小了问题的规模; 可以把计算平台机群中的任何 一台计算机上计 算出的当前全局最佳本分值, 实时地广播给所有其他并行的计算机,并作为它们新的最佳本分值, 实 现分 节点的快速并行淘汰; 应用启发式算法修改了分 定界算法, 提高了分 节点的淘汰效率选 用旅行商问题实例作为测试基准计算表明,在保证求得最优解的前提下, 该平台能很好地提高分 定界算法的效率 :分 定界算法;分布并行计算; 启发式算法;旅行商问题(TSP) : TP301 : R esearch on distribu ted parallel com pu ting for branch and bound algorithm LIY im ing, LI Y,i ZHOU M ingtian (Schoo l of C omp u ter S cience and Eng ineering, University of E lectronic Science and Technology of Ch ina, Chengdu S ichuan 610054, Ch ina) A bstract: The realization of a special cluster for calculating branch and bound algorithm was presented. The distributed parallel strategy in the cluster decreased the branch and bound running tmi e from three aspects: ( 1) the problems scalewas decreased by using distributed parallel computing; ( 2) the present optmi alvaluewhichwas calculated by any computer could be broadcast to every computer in the platform mi mediately; ( 3) the changed branch and bound algorithm with the heuristic algorithm in the cluster increases the efficiency of elmi inating node. The solving TSP(Traveling Salesman Problem) examples as benchmark show s that this platform can increase the branch and bound algorithms efficiency and get the optmi al solution sooner. K ey w ords: branch and bound algorithm; distributed parallel computing; heuristic algorithm; traveling salesman problem , 0 引