差分进化算法的交叉概率因子递增策略研究

Study on strategy of increasing cross rate in differential evolution algorithm

为了有效地控制差分进化算法的全局搜索和局部搜索,基于递增交叉概率因子的基本思想,在已有的自适应二次变异差分进化算法的基础上,提出了开口向下抛物线、开口向上抛物线和指数曲线3种非线性的交叉概率因子递增策略,并用4种经典函数对它进行测试。测试结果表明,对于多数连续优化问题,在初始参数相同的情况下,凹函数递增策略优于线性策略,而线性优于凸函数策略。凹函数递增策略能够在不影响收敛精度的情况下较大幅度地提高差分进化算法的收敛速度。

To efficiently control the global and local search of Differential Evolution algorithm（DE）,motivated by the idea of in creasing Cross Rate（CR）,three nonlinear strategies for CR,a parabola opening upwards,a parabola opening downwards and an exponential curve,based on the existing differential evolntion algorithm with adaptive second mutation are proposed.Four classic Benchmarks functions are used to evaluate the strategies on the DE performance.The experimental results show that for most continuous optimization problems,the strategy of concave function gains an advantage over the linear strategy,while the linear strategy outperforms strategy of convex function with the identical initial and final weights.