具有多种自适应机制的差分进化算法研究

Research on Differential Evolution Algorithm With Multiple Self-adapting Mechanisms

差分进化算法(DE)是一种较新的进化计算技术,具有结构简单、便于编程、求解收敛快速等优点,得到了广泛的关注和应用。为了解决经典DE计算开销大,参数设置与问题本身过于相关等缺陷,提出了一种具有多种自适应机制的改进差分进化算法(MSDE),它采用了一种自适应变异算子,可根据进化代数实时地调整变异步长,从而提高算法的求解精度,同时采用一种自适应控制参数机制,以加快算法收敛,提高算法求解成功率。通过在MATLAB仿真环境下对著名的基准测试函数分别进行求解,将改进后的算法和已有的多种优化算法进行比较,结果表明,改进的MSDE算法性能明显优于已知的算法,证明自适应是一种有效的改进思路。

Differential Evolution(DE)is a novel evolutionary computation technique,which has attracted much attention and wide applications for its simple concept,easy implementation and quick convergence.In order to tackle much overhead,problem-dependent parameters,etc and enhance the precision of classical DE,a Multiple Self-adapting DE(MSDE)algorithm is proposed by using an dynamical mutation operator adjusting the step size with evolution and a self-adapting mechanism to adjust parameters to improve the convergence and robustness.Experiments of solving well-known benchmark functions in MATLAB show the improved approach outperforms existing algorithms,and self-adapting mechanisms are effective improvement ideas.