基于惩罚函数的母子空间不同分辨率遗传算法研究

Research on Multi-resolution Genetic Algorithm Based on Penalty Function of Mother &- Son Spaces

提出一种基于惩罚函数的母子空间不同分辨率搜索的遗传算法:在母空间中,该算法采用粗分辨率搜索,待搜索速度变慢后,用特征空间识别法优化各聚类中心,并在各聚类中心点创建子空间;在子空间中,因引入惩罚函数,较好处理了子空间边界的归一化问题,并采用细分辨率搜索,以精确获得该子空间中的最优解,从而解决了传统遗传算法搜索效率和寻找全局最优解之间的矛盾。最后,通过实例验证上述算法的可行性及有效性。

A kind of multi-resolution genetic algorithm based on penalty function of mother &. son spaces is proposed in order to overcome the contradictory between search efficiency and global optimization of classical genetic algorithm. This algorithm searches mother-space with low resolution, after the search speed becomes slow, the focus center is optimized using the method of characteristic identification and the son space is established on the focus center. In son space, the penalty function is introduced to solve normalized problem of son space boundary, and an optimal solution of the son space is got by means of searching this son space with high resolution. Finally, practical results show the efficiency and effectiveness of this algorithm.