变焦遗传算法及其并行实现研究

ON ZOOMING GENETIC ALGORITHM AND ITS PARALLEL REALIZATION

遗传算法（ＧｅｎｅｔｉｃＡｌｏｇｒｉｔｈｍ，ＧＡ）以其简洁及适应性而获得广泛应用．通常遗传算法由于串格式及串长的限制，搜索空间及分辨率是有限的，因而往往收敛于局部最优．文中提出了适宜串、并行计算的变焦（Ｚｏｏｍｉｎｇ）遗传算法及多任务并行策略，采用了解码因子、搜索中心、快速变异等策略来解决搜索空间与分辨率的矛盾，并在工作站及ｔｒａｎｓｐｕｔｅｒ上分别以串行。

Genetic algorithm (GA) is widely applied to many fields due to its simplicity and adaptability. Generally, the search space and resolution of GA are limited by the string of GA, therefore, the solutions given by GA are usually local results. This paper analyzes the problems in GA and proposes the zooming GA(ZGA), which is suited to sequential and parallel computation. Strategies such as decoded factor, search center, fast mutation, etc. are adopted and the ZGA is realized on a wokrstation and transputers in sequential and parallel manners, respectively.