论文部分内容阅读
针对遗传算法的早熟现象、模式欺骗和局部搜索能力弱等缺陷提出了一个协同进化计算模型。该模型在基本遗传算法基础上引入多种群协同进化,利用性别特征改进交叉算子,采用小生境技术保持多样性。通过将问题求解转变为有向图遍历提出了染色体基因定位算子,使得问题求解和染色体中基因定位结合起来,提高问题求解的质量。在定位过程中可以和模拟退火等局部搜索能力强的启发搜索算法结合。通过几个非常容易陷入局部最优的测试函数和多峰函数测试,表明几乎所有的峰值都得到了理论值。
Aiming at the defects of the premature phenomenon of GA, mode deception and weak local search ability, a co-evolutionary computation model is proposed. Based on the basic genetic algorithm, this model introduces multi-species co-evolution, uses crossover operator to improve the crossover operator and adopts niche technology to maintain the diversity. By solving the problem solving into the directed graph traversal, a chromosome locating operator is proposed, which makes the problem solving and the gene locating in the chromosome combine to improve the quality of problem solving. It can be combined with the heuristic search algorithm with strong local search capability such as simulated annealing in the localization process. Through several test functions and multi-peak function tests, which are very easy to fall into the local optimum, it shows that almost all of the peaks have got the theoretical value.