动态种群差分进化算法研究

Research on Dynamic Population Differential Evolution Algorithm

针对标准差分进化算法易早熟的缺点,模拟人类社会民族融合的进化历程,提出了动态种群差分进化算法(DPDE)。算法中将种群分为多个独立的子种群,子种群之间采用相互移民来进行信息交换,设置种群分裂和融合的条件来动态控制子种群个数。通过数值实验用几种典型的测试函数对DPDE的搜索性能进行了测试,实验结果表明,该算法能有效地避免早熟,具有良好的全局收敛性。

Since the standard differential evolution is easy to fall into premature convergence, the paper presented a Dynamic Population Differential Evolution Algorithm （DPDE） which was based on simulating the evolution developing history of human races. The population was divided into several subpopulations in DPDE, and the numbers of subpopulations was dynamically controlled by merging and dividing, which could enhance the searching capacity of the algorithm. Finally, three experiments were made on benchmark functions. The results show that DPDE has a good global searching capacity than DE and can avoid premature convergence.