一种改进的动力学演化算法  

An Improved Dynamical Evolutionary Algorithm

作　　者：肖若辉[1] 胡豪[2] 

机构地区：[1]温州科技职业学院计算机系,浙江温州325006 [2]江苏工业学院信息科学与工程学院,江苏常州213164

出　　处：《计算机仿真》2009年第7期204-207,共4页Computer Simulation

摘　　要：动力学演化算法(DEA)是一种新颖的基于统计机制理论的演化计算技术。DEA通过驱动所有的个体运动和演化,可以有效地保持种群的多样性,但是在解决一些困难的函数优化问题时,DEA收敛速度慢并且易收敛于局部最优解。提出了一种改进的带有多父体杂交和差分变异算子的动力学演化算法(IDEA),有效地加快了DEA的收敛速度并且可以轻易逃离次优解。通过解决典型的数值函数优化问题来证实算法的有效性,实验结果表明,改进的动力学演化算法具有更高的收敛速度和收敛精度。The dynamical evolutionary algorithm （DEA） is a novel evolutionary computation technology, based on the theory of statistical mechanics. DEA effectively maintains population diversity by driving all individuals to move and to evolve. However, DEA converges slowly and often unexpectedly inclines to converge at local optima in hard function optimization. In this paper, an improved dynamical evolutionary algorithm （IDEA） with multi -parent crossover and differential evolution mutation is proposed for accelerating convergence velocity and easily escaping suboptimal solutions. In order to confirm the effectiveness of this algorithm, IDEA is applied to solve the typical numerical function minimization problems. The experimental results show that IDEA outperforms the DEA in the aspect of convergence velocity and precision.

关 键 词：动力学演化算法 统计机制 多父体杂交 差分进化 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]